|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
|
|
|
//
|
|
|
|
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
// found in the LICENSE file. See the AUTHORS file for names of contributors.
|
|
|
|
|
|
|
|
#include "db/db_iter.h"
|
|
|
|
#include <string>
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
#include <iostream>
|
|
|
|
#include <limits>
|
|
|
|
|
|
|
|
#include "db/dbformat.h"
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
#include "db/merge_context.h"
|
|
|
|
#include "db/merge_helper.h"
|
|
|
|
#include "db/pinned_iterators_manager.h"
|
|
|
|
#include "monitoring/perf_context_imp.h"
|
|
|
|
#include "rocksdb/env.h"
|
|
|
|
#include "rocksdb/iterator.h"
|
|
|
|
#include "rocksdb/merge_operator.h"
|
|
|
|
#include "rocksdb/options.h"
|
|
|
|
#include "table/internal_iterator.h"
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
#include "util/arena.h"
|
|
|
|
#include "util/filename.h"
|
|
|
|
#include "util/logging.h"
|
|
|
|
#include "util/mutexlock.h"
|
|
|
|
#include "util/string_util.h"
|
|
|
|
#include "util/trace_replay.h"
|
|
|
|
#include "util/user_comparator_wrapper.h"
|
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
|
|
|
#if 0
|
|
|
|
static void DumpInternalIter(Iterator* iter) {
|
|
|
|
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
|
|
|
|
ParsedInternalKey k;
|
|
|
|
if (!ParseInternalKey(iter->key(), &k)) {
|
|
|
|
fprintf(stderr, "Corrupt '%s'\n", EscapeString(iter->key()).c_str());
|
|
|
|
} else {
|
|
|
|
fprintf(stderr, "@ '%s'\n", k.DebugString().c_str());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
// Memtables and sstables that make the DB representation contain
|
|
|
|
// (userkey,seq,type) => uservalue entries. DBIter
|
|
|
|
// combines multiple entries for the same userkey found in the DB
|
|
|
|
// representation into a single entry while accounting for sequence
|
|
|
|
// numbers, deletion markers, overwrites, etc.
|
|
|
|
class DBIter final: public Iterator {
|
|
|
|
public:
|
|
|
|
// The following is grossly complicated. TODO: clean it up
|
|
|
|
// Which direction is the iterator currently moving?
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// (1) When moving forward:
|
|
|
|
// (1a) if current_entry_is_merged_ = false, the internal iterator is
|
|
|
|
// positioned at the exact entry that yields this->key(), this->value()
|
|
|
|
// (1b) if current_entry_is_merged_ = true, the internal iterator is
|
|
|
|
// positioned immediately after the last entry that contributed to the
|
|
|
|
// current this->value(). That entry may or may not have key equal to
|
|
|
|
// this->key().
|
|
|
|
// (2) When moving backwards, the internal iterator is positioned
|
|
|
|
// just before all entries whose user key == this->key().
|
|
|
|
enum Direction {
|
|
|
|
kForward,
|
|
|
|
kReverse
|
|
|
|
};
|
|
|
|
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
// LocalStatistics contain Statistics counters that will be aggregated per
|
|
|
|
// each iterator instance and then will be sent to the global statistics when
|
|
|
|
// the iterator is destroyed.
|
|
|
|
//
|
|
|
|
// The purpose of this approach is to avoid perf regression happening
|
|
|
|
// when multiple threads bump the atomic counters from a DBIter::Next().
|
|
|
|
struct LocalStatistics {
|
|
|
|
explicit LocalStatistics() { ResetCounters(); }
|
|
|
|
|
|
|
|
void ResetCounters() {
|
|
|
|
next_count_ = 0;
|
|
|
|
next_found_count_ = 0;
|
|
|
|
prev_count_ = 0;
|
|
|
|
prev_found_count_ = 0;
|
|
|
|
bytes_read_ = 0;
|
|
|
|
skip_count_ = 0;
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
}
|
|
|
|
|
|
|
|
void BumpGlobalStatistics(Statistics* global_statistics) {
|
|
|
|
RecordTick(global_statistics, NUMBER_DB_NEXT, next_count_);
|
|
|
|
RecordTick(global_statistics, NUMBER_DB_NEXT_FOUND, next_found_count_);
|
|
|
|
RecordTick(global_statistics, NUMBER_DB_PREV, prev_count_);
|
|
|
|
RecordTick(global_statistics, NUMBER_DB_PREV_FOUND, prev_found_count_);
|
|
|
|
RecordTick(global_statistics, ITER_BYTES_READ, bytes_read_);
|
|
|
|
RecordTick(global_statistics, NUMBER_ITER_SKIP, skip_count_);
|
|
|
|
PERF_COUNTER_ADD(iter_read_bytes, bytes_read_);
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
ResetCounters();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Map to Tickers::NUMBER_DB_NEXT
|
|
|
|
uint64_t next_count_;
|
|
|
|
// Map to Tickers::NUMBER_DB_NEXT_FOUND
|
|
|
|
uint64_t next_found_count_;
|
|
|
|
// Map to Tickers::NUMBER_DB_PREV
|
|
|
|
uint64_t prev_count_;
|
|
|
|
// Map to Tickers::NUMBER_DB_PREV_FOUND
|
|
|
|
uint64_t prev_found_count_;
|
|
|
|
// Map to Tickers::ITER_BYTES_READ
|
|
|
|
uint64_t bytes_read_;
|
|
|
|
// Map to Tickers::NUMBER_ITER_SKIP
|
|
|
|
uint64_t skip_count_;
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
};
|
|
|
|
|
|
|
|
DBIter(Env* _env, const ReadOptions& read_options,
|
|
|
|
const ImmutableCFOptions& cf_options,
|
|
|
|
const MutableCFOptions& mutable_cf_options, const Comparator* cmp,
|
|
|
|
InternalIterator* iter, SequenceNumber s, bool arena_mode,
|
|
|
|
uint64_t max_sequential_skip_in_iterations,
|
|
|
|
ReadCallback* read_callback, DBImpl* db_impl, ColumnFamilyData* cfd,
|
|
|
|
bool allow_blob)
|
|
|
|
: env_(_env),
|
|
|
|
logger_(cf_options.info_log),
|
|
|
|
user_comparator_(cmp),
|
|
|
|
merge_operator_(cf_options.merge_operator),
|
|
|
|
iter_(iter),
|
|
|
|
read_callback_(read_callback),
|
|
|
|
sequence_(s),
|
|
|
|
statistics_(cf_options.statistics),
|
|
|
|
num_internal_keys_skipped_(0),
|
|
|
|
iterate_lower_bound_(read_options.iterate_lower_bound),
|
|
|
|
iterate_upper_bound_(read_options.iterate_upper_bound),
|
|
|
|
direction_(kForward),
|
|
|
|
valid_(false),
|
|
|
|
current_entry_is_merged_(false),
|
|
|
|
prefix_same_as_start_(read_options.prefix_same_as_start),
|
|
|
|
pin_thru_lifetime_(read_options.pin_data),
|
|
|
|
total_order_seek_(read_options.total_order_seek),
|
|
|
|
allow_blob_(allow_blob),
|
|
|
|
is_blob_(false),
|
|
|
|
arena_mode_(arena_mode),
|
|
|
|
range_del_agg_(&cf_options.internal_comparator, s),
|
|
|
|
db_impl_(db_impl),
|
|
|
|
cfd_(cfd),
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
start_seqnum_(read_options.iter_start_seqnum) {
|
|
|
|
RecordTick(statistics_, NO_ITERATOR_CREATED);
|
|
|
|
prefix_extractor_ = mutable_cf_options.prefix_extractor.get();
|
|
|
|
max_skip_ = max_sequential_skip_in_iterations;
|
|
|
|
max_skippable_internal_keys_ = read_options.max_skippable_internal_keys;
|
|
|
|
if (pin_thru_lifetime_) {
|
|
|
|
pinned_iters_mgr_.StartPinning();
|
|
|
|
}
|
|
|
|
if (iter_) {
|
|
|
|
iter_->SetPinnedItersMgr(&pinned_iters_mgr_);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
~DBIter() override {
|
|
|
|
// Release pinned data if any
|
|
|
|
if (pinned_iters_mgr_.PinningEnabled()) {
|
|
|
|
pinned_iters_mgr_.ReleasePinnedData();
|
|
|
|
}
|
|
|
|
RecordTick(statistics_, NO_ITERATOR_DELETED);
|
|
|
|
ResetInternalKeysSkippedCounter();
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
local_stats_.BumpGlobalStatistics(statistics_);
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
if (!arena_mode_) {
|
|
|
|
delete iter_;
|
|
|
|
} else {
|
|
|
|
iter_->~InternalIterator();
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
}
|
|
|
|
}
|
|
|
|
virtual void SetIter(InternalIterator* iter) {
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
assert(iter_ == nullptr);
|
|
|
|
iter_ = iter;
|
|
|
|
iter_->SetPinnedItersMgr(&pinned_iters_mgr_);
|
|
|
|
}
|
|
|
|
virtual ReadRangeDelAggregator* GetRangeDelAggregator() {
|
|
|
|
return &range_del_agg_;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool Valid() const override { return valid_; }
|
|
|
|
Slice key() const override {
|
|
|
|
assert(valid_);
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
if(start_seqnum_ > 0) {
|
|
|
|
return saved_key_.GetInternalKey();
|
|
|
|
} else {
|
|
|
|
return saved_key_.GetUserKey();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
Slice value() const override {
|
|
|
|
assert(valid_);
|
|
|
|
if (current_entry_is_merged_) {
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
// If pinned_value_ is set then the result of merge operator is one of
|
|
|
|
// the merge operands and we should return it.
|
|
|
|
return pinned_value_.data() ? pinned_value_ : saved_value_;
|
|
|
|
} else if (direction_ == kReverse) {
|
|
|
|
return pinned_value_;
|
|
|
|
} else {
|
|
|
|
return iter_->value();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
Status status() const override {
|
|
|
|
if (status_.ok()) {
|
|
|
|
return iter_->status();
|
|
|
|
} else {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
assert(!valid_);
|
|
|
|
return status_;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
bool IsBlob() const {
|
|
|
|
assert(valid_ && (allow_blob_ || !is_blob_));
|
|
|
|
return is_blob_;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status GetProperty(std::string prop_name, std::string* prop) override {
|
|
|
|
if (prop == nullptr) {
|
|
|
|
return Status::InvalidArgument("prop is nullptr");
|
|
|
|
}
|
|
|
|
if (prop_name == "rocksdb.iterator.super-version-number") {
|
|
|
|
// First try to pass the value returned from inner iterator.
|
|
|
|
return iter_->GetProperty(prop_name, prop);
|
|
|
|
} else if (prop_name == "rocksdb.iterator.is-key-pinned") {
|
|
|
|
if (valid_) {
|
|
|
|
*prop = (pin_thru_lifetime_ && saved_key_.IsKeyPinned()) ? "1" : "0";
|
|
|
|
} else {
|
|
|
|
*prop = "Iterator is not valid.";
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
} else if (prop_name == "rocksdb.iterator.internal-key") {
|
|
|
|
*prop = saved_key_.GetUserKey().ToString();
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
return Status::InvalidArgument("Unidentified property.");
|
|
|
|
}
|
|
|
|
|
|
|
|
void Next() final override;
|
|
|
|
void Prev() override;
|
|
|
|
void Seek(const Slice& target) override;
|
|
|
|
void SeekForPrev(const Slice& target) override;
|
|
|
|
void SeekToFirst() override;
|
|
|
|
void SeekToLast() override;
|
|
|
|
Env* env() { return env_; }
|
|
|
|
void set_sequence(uint64_t s) {
|
|
|
|
sequence_ = s;
|
|
|
|
if (read_callback_) {
|
|
|
|
read_callback_->Refresh(s);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
void set_valid(bool v) { valid_ = v; }
|
|
|
|
|
|
|
|
private:
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// For all methods in this block:
|
|
|
|
// PRE: iter_->Valid() && status_.ok()
|
|
|
|
// Return false if there was an error, and status() is non-ok, valid_ = false;
|
|
|
|
// in this case callers would usually stop what they were doing and return.
|
|
|
|
bool ReverseToForward();
|
|
|
|
bool ReverseToBackward();
|
|
|
|
bool FindValueForCurrentKey();
|
|
|
|
bool FindValueForCurrentKeyUsingSeek();
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
bool FindUserKeyBeforeSavedKey();
|
|
|
|
inline bool FindNextUserEntry(bool skipping, bool prefix_check);
|
|
|
|
inline bool FindNextUserEntryInternal(bool skipping, bool prefix_check);
|
|
|
|
bool ParseKey(ParsedInternalKey* key);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
bool MergeValuesNewToOld();
|
|
|
|
|
|
|
|
void PrevInternal();
|
|
|
|
bool TooManyInternalKeysSkipped(bool increment = true);
|
|
|
|
inline bool IsVisible(SequenceNumber sequence);
|
|
|
|
|
|
|
|
// CanReseekToSkip() returns whether the iterator can use the optimization
|
|
|
|
// where it reseek by sequence number to get the next key when there are too
|
|
|
|
// many versions. This is disabled for write unprepared because seeking to
|
|
|
|
// sequence number does not guarantee that it is visible.
|
|
|
|
inline bool CanReseekToSkip();
|
|
|
|
|
|
|
|
// Temporarily pin the blocks that we encounter until ReleaseTempPinnedData()
|
|
|
|
// is called
|
|
|
|
void TempPinData() {
|
|
|
|
if (!pin_thru_lifetime_) {
|
|
|
|
pinned_iters_mgr_.StartPinning();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Release blocks pinned by TempPinData()
|
|
|
|
void ReleaseTempPinnedData() {
|
|
|
|
if (!pin_thru_lifetime_ && pinned_iters_mgr_.PinningEnabled()) {
|
|
|
|
pinned_iters_mgr_.ReleasePinnedData();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
inline void ClearSavedValue() {
|
|
|
|
if (saved_value_.capacity() > 1048576) {
|
|
|
|
std::string empty;
|
|
|
|
swap(empty, saved_value_);
|
|
|
|
} else {
|
|
|
|
saved_value_.clear();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
inline void ResetInternalKeysSkippedCounter() {
|
|
|
|
local_stats_.skip_count_ += num_internal_keys_skipped_;
|
|
|
|
if (valid_) {
|
|
|
|
local_stats_.skip_count_--;
|
|
|
|
}
|
|
|
|
num_internal_keys_skipped_ = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
const SliceTransform* prefix_extractor_;
|
|
|
|
Env* const env_;
|
|
|
|
Logger* logger_;
|
|
|
|
UserComparatorWrapper user_comparator_;
|
|
|
|
const MergeOperator* const merge_operator_;
|
|
|
|
InternalIterator* iter_;
|
|
|
|
ReadCallback* read_callback_;
|
|
|
|
// Max visible sequence number. It is normally the snapshot seq unless we have
|
|
|
|
// uncommitted data in db as in WriteUnCommitted.
|
|
|
|
SequenceNumber sequence_;
|
|
|
|
|
|
|
|
IterKey saved_key_;
|
|
|
|
// Reusable internal key data structure. This is only used inside one function
|
|
|
|
// and should not be used across functions. Reusing this object can reduce
|
|
|
|
// overhead of calling construction of the function if creating it each time.
|
|
|
|
ParsedInternalKey ikey_;
|
|
|
|
std::string saved_value_;
|
|
|
|
Slice pinned_value_;
|
|
|
|
// for prefix seek mode to support prev()
|
|
|
|
Statistics* statistics_;
|
|
|
|
uint64_t max_skip_;
|
|
|
|
uint64_t max_skippable_internal_keys_;
|
|
|
|
uint64_t num_internal_keys_skipped_;
|
|
|
|
const Slice* iterate_lower_bound_;
|
|
|
|
const Slice* iterate_upper_bound_;
|
|
|
|
|
|
|
|
IterKey prefix_start_buf_;
|
|
|
|
|
|
|
|
Status status_;
|
|
|
|
Slice prefix_start_key_;
|
|
|
|
Direction direction_;
|
|
|
|
bool valid_;
|
|
|
|
bool current_entry_is_merged_;
|
|
|
|
const bool prefix_same_as_start_;
|
|
|
|
// Means that we will pin all data blocks we read as long the Iterator
|
|
|
|
// is not deleted, will be true if ReadOptions::pin_data is true
|
|
|
|
const bool pin_thru_lifetime_;
|
|
|
|
const bool total_order_seek_;
|
|
|
|
bool allow_blob_;
|
|
|
|
bool is_blob_;
|
|
|
|
bool arena_mode_;
|
|
|
|
// List of operands for merge operator.
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
MergeContext merge_context_;
|
|
|
|
ReadRangeDelAggregator range_del_agg_;
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
LocalStatistics local_stats_;
|
|
|
|
PinnedIteratorsManager pinned_iters_mgr_;
|
|
|
|
DBImpl* db_impl_;
|
|
|
|
ColumnFamilyData* cfd_;
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
// for diff snapshots we want the lower bound on the seqnum;
|
|
|
|
// if this value > 0 iterator will return internal keys
|
|
|
|
SequenceNumber start_seqnum_;
|
|
|
|
|
|
|
|
// No copying allowed
|
|
|
|
DBIter(const DBIter&);
|
|
|
|
void operator=(const DBIter&);
|
|
|
|
};
|
|
|
|
|
|
|
|
inline bool DBIter::ParseKey(ParsedInternalKey* ikey) {
|
|
|
|
if (!ParseInternalKey(iter_->key(), ikey)) {
|
|
|
|
status_ = Status::Corruption("corrupted internal key in DBIter");
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
valid_ = false;
|
|
|
|
ROCKS_LOG_ERROR(logger_, "corrupted internal key in DBIter: %s",
|
|
|
|
iter_->key().ToString(true).c_str());
|
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::Next() {
|
|
|
|
assert(valid_);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
assert(status_.ok());
|
|
|
|
|
|
|
|
PERF_CPU_TIMER_GUARD(iter_next_cpu_nanos, env_);
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
// Release temporarily pinned blocks from last operation
|
|
|
|
ReleaseTempPinnedData();
|
|
|
|
local_stats_.skip_count_ += num_internal_keys_skipped_;
|
|
|
|
local_stats_.skip_count_--;
|
|
|
|
num_internal_keys_skipped_ = 0;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
bool ok = true;
|
|
|
|
if (direction_ == kReverse) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!ReverseToForward()) {
|
|
|
|
ok = false;
|
|
|
|
}
|
|
|
|
} else if (!current_entry_is_merged_) {
|
|
|
|
// If the current value is not a merge, the iter position is the
|
|
|
|
// current key, which is already returned. We can safely issue a
|
|
|
|
// Next() without checking the current key.
|
|
|
|
// If the current key is a merge, very likely iter already points
|
|
|
|
// to the next internal position.
|
|
|
|
assert(iter_->Valid());
|
|
|
|
iter_->Next();
|
|
|
|
PERF_COUNTER_ADD(internal_key_skipped_count, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
local_stats_.next_count_++;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (ok && iter_->Valid()) {
|
|
|
|
FindNextUserEntry(true /* skipping the current user key */,
|
|
|
|
prefix_same_as_start_);
|
|
|
|
} else {
|
|
|
|
valid_ = false;
|
|
|
|
}
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
if (statistics_ != nullptr && valid_) {
|
|
|
|
local_stats_.next_found_count_++;
|
|
|
|
local_stats_.bytes_read_ += (key().size() + value().size());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// PRE: saved_key_ has the current user key if skipping
|
|
|
|
// POST: saved_key_ should have the next user key if valid_,
|
|
|
|
// if the current entry is a result of merge
|
|
|
|
// current_entry_is_merged_ => true
|
|
|
|
// saved_value_ => the merged value
|
|
|
|
//
|
|
|
|
// NOTE: In between, saved_key_ can point to a user key that has
|
|
|
|
// a delete marker or a sequence number higher than sequence_
|
|
|
|
// saved_key_ MUST have a proper user_key before calling this function
|
|
|
|
//
|
|
|
|
// The prefix_check parameter controls whether we check the iterated
|
|
|
|
// keys against the prefix of the seeked key. Set to false when
|
|
|
|
// performing a seek without a key (e.g. SeekToFirst). Set to
|
|
|
|
// prefix_same_as_start_ for other iterations.
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
inline bool DBIter::FindNextUserEntry(bool skipping, bool prefix_check) {
|
|
|
|
PERF_TIMER_GUARD(find_next_user_entry_time);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return FindNextUserEntryInternal(skipping, prefix_check);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Actual implementation of DBIter::FindNextUserEntry()
|
|
|
|
inline bool DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check) {
|
|
|
|
// Loop until we hit an acceptable entry to yield
|
|
|
|
assert(iter_->Valid());
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
assert(status_.ok());
|
|
|
|
assert(direction_ == kForward);
|
|
|
|
current_entry_is_merged_ = false;
|
|
|
|
|
|
|
|
// How many times in a row we have skipped an entry with user key less than
|
|
|
|
// or equal to saved_key_. We could skip these entries either because
|
|
|
|
// sequence numbers were too high or because skipping = true.
|
|
|
|
// What saved_key_ contains throughout this method:
|
|
|
|
// - if skipping : saved_key_ contains the key that we need to skip,
|
|
|
|
// and we haven't seen any keys greater than that,
|
|
|
|
// - if num_skipped > 0 : saved_key_ contains the key that we have skipped
|
|
|
|
// num_skipped times, and we haven't seen any keys
|
|
|
|
// greater than that,
|
|
|
|
// - none of the above : saved_key_ can contain anything, it doesn't matter.
|
|
|
|
uint64_t num_skipped = 0;
|
|
|
|
|
|
|
|
is_blob_ = false;
|
|
|
|
|
|
|
|
do {
|
|
|
|
if (!ParseKey(&ikey_)) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (iterate_upper_bound_ != nullptr &&
|
|
|
|
user_comparator_.Compare(ikey_.user_key, *iterate_upper_bound_) >= 0) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (prefix_extractor_ && prefix_check &&
|
|
|
|
prefix_extractor_->Transform(ikey_.user_key)
|
|
|
|
.compare(prefix_start_key_) != 0) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (TooManyInternalKeysSkipped()) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (IsVisible(ikey_.sequence)) {
|
|
|
|
if (skipping && user_comparator_.Compare(ikey_.user_key,
|
|
|
|
saved_key_.GetUserKey()) <= 0) {
|
|
|
|
num_skipped++; // skip this entry
|
|
|
|
PERF_COUNTER_ADD(internal_key_skipped_count, 1);
|
|
|
|
} else {
|
|
|
|
num_skipped = 0;
|
|
|
|
switch (ikey_.type) {
|
|
|
|
case kTypeDeletion:
|
|
|
|
case kTypeSingleDeletion:
|
|
|
|
// Arrange to skip all upcoming entries for this key since
|
|
|
|
// they are hidden by this deletion.
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
// if iterartor specified start_seqnum we
|
|
|
|
// 1) return internal key, including the type
|
|
|
|
// 2) return ikey only if ikey.seqnum >= start_seqnum_
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// note that if deletion seqnum is < start_seqnum_ we
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
// just skip it like in normal iterator.
|
|
|
|
if (start_seqnum_ > 0 && ikey_.sequence >= start_seqnum_) {
|
|
|
|
saved_key_.SetInternalKey(ikey_);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
valid_ = true;
|
|
|
|
return true;
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
} else {
|
|
|
|
saved_key_.SetUserKey(
|
|
|
|
ikey_.user_key,
|
|
|
|
!pin_thru_lifetime_ || !iter_->IsKeyPinned() /* copy */);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
skipping = true;
|
|
|
|
PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
}
|
|
|
|
break;
|
|
|
|
case kTypeValue:
|
|
|
|
case kTypeBlobIndex:
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
if (start_seqnum_ > 0) {
|
|
|
|
// we are taking incremental snapshot here
|
|
|
|
// incremental snapshots aren't supported on DB with range deletes
|
|
|
|
assert(!(
|
|
|
|
(ikey_.type == kTypeBlobIndex) && (start_seqnum_ > 0)
|
|
|
|
));
|
|
|
|
if (ikey_.sequence >= start_seqnum_) {
|
|
|
|
saved_key_.SetInternalKey(ikey_);
|
|
|
|
valid_ = true;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
} else {
|
|
|
|
// this key and all previous versions shouldn't be included,
|
|
|
|
// skipping
|
|
|
|
saved_key_.SetUserKey(ikey_.user_key,
|
|
|
|
!pin_thru_lifetime_ || !iter_->IsKeyPinned() /* copy */);
|
|
|
|
skipping = true;
|
|
|
|
}
|
|
|
|
} else {
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
saved_key_.SetUserKey(
|
|
|
|
ikey_.user_key,
|
|
|
|
!pin_thru_lifetime_ || !iter_->IsKeyPinned() /* copy */);
|
|
|
|
if (range_del_agg_.ShouldDelete(
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
ikey_, RangeDelPositioningMode::kForwardTraversal)) {
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
// Arrange to skip all upcoming entries for this key since
|
|
|
|
// they are hidden by this deletion.
|
|
|
|
skipping = true;
|
|
|
|
num_skipped = 0;
|
|
|
|
PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
|
|
|
|
} else if (ikey_.type == kTypeBlobIndex) {
|
|
|
|
if (!allow_blob_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
|
|
|
|
status_ = Status::NotSupported(
|
|
|
|
"Encounter unexpected blob index. Please open DB with "
|
|
|
|
"rocksdb::blob_db::BlobDB instead.");
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
is_blob_ = true;
|
|
|
|
valid_ = true;
|
|
|
|
return true;
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
} else {
|
|
|
|
valid_ = true;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
Added support for differential snapshots
Summary:
The motivation for this PR is to add to RocksDB support for differential (incremental) snapshots, as snapshot of the DB changes between two points in time (one can think of it as diff between to sequence numbers, or the diff D which can be thought of as an SST file or just set of KVs that can be applied to sequence number S1 to get the database to the state at sequence number S2).
This feature would be useful for various distributed storages layers built on top of RocksDB, as it should help reduce resources (time and network bandwidth) needed to recover and rebuilt DB instances as replicas in the context of distributed storages.
From the API standpoint that would like client app requesting iterator between (start seqnum) and current DB state, and reading the "diff".
This is a very draft PR for initial review in the discussion on the approach, i'm going to rework some parts and keep updating the PR.
For now, what's done here according to initial discussions:
Preserving deletes:
- We want to be able to optionally preserve recent deletes for some defined period of time, so that if a delete came in recently and might need to be included in the next incremental snapshot it would't get dropped by a compaction. This is done by adding new param to Options (preserve deletes flag) and new variable to DB Impl where we keep track of the sequence number after which we don't want to drop tombstones, even if they are otherwise eligible for deletion.
- I also added a new API call for clients to be able to advance this cutoff seqnum after which we drop deletes; i assume it's more flexible to let clients control this, since otherwise we'd need to keep some kind of timestamp < -- > seqnum mapping inside the DB, which sounds messy and painful to support. Clients could make use of it by periodically calling GetLatestSequenceNumber(), noting the timestamp, doing some calculation and figuring out by how much we need to advance the cutoff seqnum.
- Compaction codepath in compaction_iterator.cc has been modified to avoid dropping tombstones with seqnum > cutoff seqnum.
Iterator changes:
- couple params added to ReadOptions, to optionally allow client to request internal keys instead of user keys (so that client can get the latest value of a key, be it delete marker or a put), as well as min timestamp and min seqnum.
TableCache changes:
- I modified table_cache code to be able to quickly exclude SST files from iterators heep if creation_time on the file is less then iter_start_ts as passed in ReadOptions. That would help a lot in some DB settings (like reading very recent data only or using FIFO compactions), but not so much for universal compaction with more or less long iterator time span.
What's left:
- Still looking at how to best plug that inside DBIter codepath. So far it seems that FindNextUserKeyInternal only parses values as UserKeys, and iter->key() call generally returns user key. Can we add new API to DBIter as internal_key(), and modify this internal method to optionally set saved_key_ to point to the full internal key? I don't need to store actual seqnum there, but I do need to store type.
Closes https://github.com/facebook/rocksdb/pull/2999
Differential Revision: D6175602
Pulled By: mikhail-antonov
fbshipit-source-id: c779a6696ee2d574d86c69cec866a3ae095aa900
7 years ago
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case kTypeMerge:
|
|
|
|
saved_key_.SetUserKey(
|
|
|
|
ikey_.user_key,
|
|
|
|
!pin_thru_lifetime_ || !iter_->IsKeyPinned() /* copy */);
|
Maintain position in range deletions map
Summary:
When deletion-collapsing mode is enabled (i.e., for DBIter/CompactionIterator), we maintain position in the tombstone maps across calls to ShouldDelete(). Since iterators often access keys sequentially (or reverse-sequentially), scanning forward/backward from the last position can be faster than binary-searching the map for every key.
- When Next() is invoked on an iterator, we use kForwardTraversal to scan forwards, if needed, until arriving at the range deletion containing the next key.
- Similarly for Prev(), we use kBackwardTraversal to scan backwards in the range deletion map.
- When the iterator seeks, we use kBinarySearch for repositioning
- After tombstones are added or before the first ShouldDelete() invocation, the current position is set to invalid, which forces kBinarySearch to be used.
- Non-iterator users (i.e., Get()) use kFullScan, which has the same behavior as before---scan the whole map for every key passed to ShouldDelete().
Closes https://github.com/facebook/rocksdb/pull/1701
Differential Revision: D4350318
Pulled By: ajkr
fbshipit-source-id: 5129b76
8 years ago
|
|
|
if (range_del_agg_.ShouldDelete(
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
ikey_, RangeDelPositioningMode::kForwardTraversal)) {
|
|
|
|
// Arrange to skip all upcoming entries for this key since
|
|
|
|
// they are hidden by this deletion.
|
|
|
|
skipping = true;
|
|
|
|
num_skipped = 0;
|
|
|
|
PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
|
|
|
|
} else {
|
|
|
|
// By now, we are sure the current ikey is going to yield a
|
|
|
|
// value
|
|
|
|
current_entry_is_merged_ = true;
|
|
|
|
valid_ = true;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return MergeValuesNewToOld(); // Go to a different state machine
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
assert(false);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// This key was inserted after our snapshot was taken.
|
|
|
|
// If this happens too many times in a row for the same user key, we want
|
|
|
|
// to seek to the target sequence number.
|
|
|
|
int cmp =
|
|
|
|
user_comparator_.Compare(ikey_.user_key, saved_key_.GetUserKey());
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (cmp == 0 || (skipping && cmp <= 0)) {
|
|
|
|
num_skipped++;
|
|
|
|
} else {
|
|
|
|
saved_key_.SetUserKey(
|
|
|
|
ikey_.user_key,
|
|
|
|
!iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
|
|
|
|
skipping = false;
|
|
|
|
num_skipped = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// If we have sequentially iterated via numerous equal keys, then it's
|
|
|
|
// better to seek so that we can avoid too many key comparisons.
|
|
|
|
if (num_skipped > max_skip_ && CanReseekToSkip()) {
|
|
|
|
num_skipped = 0;
|
|
|
|
std::string last_key;
|
|
|
|
if (skipping) {
|
|
|
|
// We're looking for the next user-key but all we see are the same
|
|
|
|
// user-key with decreasing sequence numbers. Fast forward to
|
|
|
|
// sequence number 0 and type deletion (the smallest type).
|
|
|
|
AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetUserKey(),
|
|
|
|
0, kTypeDeletion));
|
|
|
|
// Don't set skipping = false because we may still see more user-keys
|
|
|
|
// equal to saved_key_.
|
|
|
|
} else {
|
|
|
|
// We saw multiple entries with this user key and sequence numbers
|
|
|
|
// higher than sequence_. Fast forward to sequence_.
|
|
|
|
// Note that this only covers a case when a higher key was overwritten
|
|
|
|
// many times since our snapshot was taken, not the case when a lot of
|
|
|
|
// different keys were inserted after our snapshot was taken.
|
|
|
|
AppendInternalKey(&last_key,
|
|
|
|
ParsedInternalKey(saved_key_.GetUserKey(), sequence_,
|
|
|
|
kValueTypeForSeek));
|
|
|
|
}
|
|
|
|
iter_->Seek(last_key);
|
|
|
|
RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
|
|
|
|
} else {
|
|
|
|
iter_->Next();
|
|
|
|
}
|
|
|
|
} while (iter_->Valid());
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return iter_->status().ok();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Merge values of the same user key starting from the current iter_ position
|
|
|
|
// Scan from the newer entries to older entries.
|
|
|
|
// PRE: iter_->key() points to the first merge type entry
|
|
|
|
// saved_key_ stores the user key
|
|
|
|
// POST: saved_value_ has the merged value for the user key
|
|
|
|
// iter_ points to the next entry (or invalid)
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
bool DBIter::MergeValuesNewToOld() {
|
|
|
|
if (!merge_operator_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Options::merge_operator is null.");
|
|
|
|
status_ = Status::InvalidArgument("merge_operator_ must be set.");
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
// Temporarily pin the blocks that hold merge operands
|
|
|
|
TempPinData();
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
merge_context_.Clear();
|
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences.
Summary:
Here are the major changes to the Merge Interface. It has been expanded
to handle cases where the MergeOperator is not associative. It does so by stacking
up merge operations while scanning through the key history (i.e.: during Get() or
Compaction), until a valid Put/Delete/end-of-history is encountered; it then
applies all of the merge operations in the correct sequence starting with the
base/sentinel value.
I have also introduced an "AssociativeMerge" function which allows the user to
take advantage of associative merge operations (such as in the case of counters).
The implementation will always attempt to merge the operations/operands themselves
together when they are encountered, and will resort to the "stacking" method if
and only if the "associative-merge" fails.
This implementation is conjectured to allow MergeOperator to handle the general
case, while still providing the user with the ability to take advantage of certain
efficiencies in their own merge-operator / data-structure.
NOTE: This is a preliminary diff. This must still go through a lot of review,
revision, and testing. Feedback welcome!
Test Plan:
-This is a preliminary diff. I have only just begun testing/debugging it.
-I will be testing this with the existing MergeOperator use-cases and unit-tests
(counters, string-append, and redis-lists)
-I will be "desk-checking" and walking through the code with the help gdb.
-I will find a way of stress-testing the new interface / implementation using
db_bench, db_test, merge_test, and/or db_stress.
-I will ensure that my tests cover all cases: Get-Memtable,
Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0,
Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found,
end-of-history, end-of-file, etc.
-A lot of feedback from the reviewers.
Reviewers: haobo, dhruba, zshao, emayanke
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D11499
11 years ago
|
|
|
// Start the merge process by pushing the first operand
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
merge_context_.PushOperand(iter_->value(),
|
|
|
|
iter_->IsValuePinned() /* operand_pinned */);
|
|
|
|
TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:PushedFirstOperand");
|
|
|
|
|
|
|
|
ParsedInternalKey ikey;
|
|
|
|
Status s;
|
|
|
|
for (iter_->Next(); iter_->Valid(); iter_->Next()) {
|
|
|
|
TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:SteppedToNextOperand");
|
|
|
|
if (!ParseKey(&ikey)) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!user_comparator_.Equal(ikey.user_key, saved_key_.GetUserKey())) {
|
|
|
|
// hit the next user key, stop right here
|
|
|
|
break;
|
|
|
|
} else if (kTypeDeletion == ikey.type || kTypeSingleDeletion == ikey.type ||
|
Maintain position in range deletions map
Summary:
When deletion-collapsing mode is enabled (i.e., for DBIter/CompactionIterator), we maintain position in the tombstone maps across calls to ShouldDelete(). Since iterators often access keys sequentially (or reverse-sequentially), scanning forward/backward from the last position can be faster than binary-searching the map for every key.
- When Next() is invoked on an iterator, we use kForwardTraversal to scan forwards, if needed, until arriving at the range deletion containing the next key.
- Similarly for Prev(), we use kBackwardTraversal to scan backwards in the range deletion map.
- When the iterator seeks, we use kBinarySearch for repositioning
- After tombstones are added or before the first ShouldDelete() invocation, the current position is set to invalid, which forces kBinarySearch to be used.
- Non-iterator users (i.e., Get()) use kFullScan, which has the same behavior as before---scan the whole map for every key passed to ShouldDelete().
Closes https://github.com/facebook/rocksdb/pull/1701
Differential Revision: D4350318
Pulled By: ajkr
fbshipit-source-id: 5129b76
8 years ago
|
|
|
range_del_agg_.ShouldDelete(
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
ikey, RangeDelPositioningMode::kForwardTraversal)) {
|
|
|
|
// hit a delete with the same user key, stop right here
|
|
|
|
// iter_ is positioned after delete
|
|
|
|
iter_->Next();
|
|
|
|
break;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
} else if (kTypeValue == ikey.type) {
|
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences.
Summary:
Here are the major changes to the Merge Interface. It has been expanded
to handle cases where the MergeOperator is not associative. It does so by stacking
up merge operations while scanning through the key history (i.e.: during Get() or
Compaction), until a valid Put/Delete/end-of-history is encountered; it then
applies all of the merge operations in the correct sequence starting with the
base/sentinel value.
I have also introduced an "AssociativeMerge" function which allows the user to
take advantage of associative merge operations (such as in the case of counters).
The implementation will always attempt to merge the operations/operands themselves
together when they are encountered, and will resort to the "stacking" method if
and only if the "associative-merge" fails.
This implementation is conjectured to allow MergeOperator to handle the general
case, while still providing the user with the ability to take advantage of certain
efficiencies in their own merge-operator / data-structure.
NOTE: This is a preliminary diff. This must still go through a lot of review,
revision, and testing. Feedback welcome!
Test Plan:
-This is a preliminary diff. I have only just begun testing/debugging it.
-I will be testing this with the existing MergeOperator use-cases and unit-tests
(counters, string-append, and redis-lists)
-I will be "desk-checking" and walking through the code with the help gdb.
-I will find a way of stress-testing the new interface / implementation using
db_bench, db_test, merge_test, and/or db_stress.
-I will ensure that my tests cover all cases: Get-Memtable,
Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0,
Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found,
end-of-history, end-of-file, etc.
-A lot of feedback from the reviewers.
Reviewers: haobo, dhruba, zshao, emayanke
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D11499
11 years ago
|
|
|
// hit a put, merge the put value with operands and store the
|
|
|
|
// final result in saved_value_. We are done!
|
|
|
|
const Slice val = iter_->value();
|
|
|
|
s = MergeHelper::TimedFullMerge(
|
|
|
|
merge_operator_, ikey.user_key, &val, merge_context_.GetOperands(),
|
|
|
|
&saved_value_, logger_, statistics_, env_, &pinned_value_, true);
|
|
|
|
if (!s.ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
status_ = s;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
// iter_ is positioned after put
|
|
|
|
iter_->Next();
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!iter_->status().ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
} else if (kTypeMerge == ikey.type) {
|
[RocksDB] [MergeOperator] The new Merge Interface! Uses merge sequences.
Summary:
Here are the major changes to the Merge Interface. It has been expanded
to handle cases where the MergeOperator is not associative. It does so by stacking
up merge operations while scanning through the key history (i.e.: during Get() or
Compaction), until a valid Put/Delete/end-of-history is encountered; it then
applies all of the merge operations in the correct sequence starting with the
base/sentinel value.
I have also introduced an "AssociativeMerge" function which allows the user to
take advantage of associative merge operations (such as in the case of counters).
The implementation will always attempt to merge the operations/operands themselves
together when they are encountered, and will resort to the "stacking" method if
and only if the "associative-merge" fails.
This implementation is conjectured to allow MergeOperator to handle the general
case, while still providing the user with the ability to take advantage of certain
efficiencies in their own merge-operator / data-structure.
NOTE: This is a preliminary diff. This must still go through a lot of review,
revision, and testing. Feedback welcome!
Test Plan:
-This is a preliminary diff. I have only just begun testing/debugging it.
-I will be testing this with the existing MergeOperator use-cases and unit-tests
(counters, string-append, and redis-lists)
-I will be "desk-checking" and walking through the code with the help gdb.
-I will find a way of stress-testing the new interface / implementation using
db_bench, db_test, merge_test, and/or db_stress.
-I will ensure that my tests cover all cases: Get-Memtable,
Get-Immutable-Memtable, Get-from-Disk, Iterator-Range-Scan, Flush-Memtable-to-L0,
Compaction-L0-L1, Compaction-Ln-L(n+1), Put/Delete found, Put/Delete not-found,
end-of-history, end-of-file, etc.
-A lot of feedback from the reviewers.
Reviewers: haobo, dhruba, zshao, emayanke
Reviewed By: haobo
CC: leveldb
Differential Revision: https://reviews.facebook.net/D11499
11 years ago
|
|
|
// hit a merge, add the value as an operand and run associative merge.
|
|
|
|
// when complete, add result to operands and continue.
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
merge_context_.PushOperand(iter_->value(),
|
|
|
|
iter_->IsValuePinned() /* operand_pinned */);
|
|
|
|
PERF_COUNTER_ADD(internal_merge_count, 1);
|
|
|
|
} else if (kTypeBlobIndex == ikey.type) {
|
|
|
|
if (!allow_blob_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
|
|
|
|
status_ = Status::NotSupported(
|
|
|
|
"Encounter unexpected blob index. Please open DB with "
|
|
|
|
"rocksdb::blob_db::BlobDB instead.");
|
|
|
|
} else {
|
|
|
|
status_ =
|
|
|
|
Status::NotSupported("Blob DB does not support merge operator.");
|
|
|
|
}
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
} else {
|
|
|
|
assert(false);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!iter_->status().ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
// we either exhausted all internal keys under this user key, or hit
|
|
|
|
// a deletion marker.
|
|
|
|
// feed null as the existing value to the merge operator, such that
|
|
|
|
// client can differentiate this scenario and do things accordingly.
|
|
|
|
s = MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetUserKey(),
|
|
|
|
nullptr, merge_context_.GetOperands(),
|
|
|
|
&saved_value_, logger_, statistics_, env_,
|
|
|
|
&pinned_value_, true);
|
|
|
|
if (!s.ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
status_ = s;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
assert(status_.ok());
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::Prev() {
|
|
|
|
assert(valid_);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
assert(status_.ok());
|
|
|
|
|
|
|
|
PERF_CPU_TIMER_GUARD(iter_prev_cpu_nanos, env_);
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
ReleaseTempPinnedData();
|
|
|
|
ResetInternalKeysSkippedCounter();
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
bool ok = true;
|
|
|
|
if (direction_ == kForward) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!ReverseToBackward()) {
|
|
|
|
ok = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (ok) {
|
|
|
|
PrevInternal();
|
|
|
|
}
|
|
|
|
if (statistics_ != nullptr) {
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
local_stats_.prev_count_++;
|
|
|
|
if (valid_) {
|
Aggregate hot Iterator counters in LocalStatistics (DBIter::Next perf regression)
Summary:
This patch bump the counters in the frequent code path DBIter::Next() / DBIter::Prev() in a local data members and send them to Statistics when the iterator is destroyed
A better solution will be to have thread_local implementation for Statistics
New performance
```
readseq : 0.035 micros/op 28597881 ops/sec; 3163.7 MB/s
1,851,568,819 stalled-cycles-frontend # 31.29% frontend cycles idle [49.86%]
884,929,823 stalled-cycles-backend # 14.95% backend cycles idle [50.21%]
readreverse : 0.071 micros/op 14077393 ops/sec; 1557.3 MB/s
3,239,575,993 stalled-cycles-frontend # 27.36% frontend cycles idle [49.96%]
1,558,253,983 stalled-cycles-backend # 13.16% backend cycles idle [50.14%]
```
Existing performance
```
readreverse : 0.174 micros/op 5732342 ops/sec; 634.1 MB/s
20,570,209,389 stalled-cycles-frontend # 70.71% frontend cycles idle [50.01%]
18,422,816,837 stalled-cycles-backend # 63.33% backend cycles idle [50.04%]
readseq : 0.119 micros/op 8400537 ops/sec; 929.3 MB/s
15,634,225,844 stalled-cycles-frontend # 79.07% frontend cycles idle [49.96%]
14,227,427,453 stalled-cycles-backend # 71.95% backend cycles idle [50.09%]
```
Test Plan: unit tests
Reviewers: yhchiang, sdong, igor
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D55107
9 years ago
|
|
|
local_stats_.prev_found_count_++;
|
|
|
|
local_stats_.bytes_read_ += (key().size() + value().size());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
bool DBIter::ReverseToForward() {
|
|
|
|
assert(iter_->status().ok());
|
|
|
|
|
|
|
|
// When moving backwards, iter_ is positioned on _previous_ key, which may
|
|
|
|
// not exist or may have different prefix than the current key().
|
|
|
|
// If that's the case, seek iter_ to current key.
|
|
|
|
if ((prefix_extractor_ != nullptr && !total_order_seek_) || !iter_->Valid()) {
|
|
|
|
IterKey last_key;
|
|
|
|
last_key.SetInternalKey(ParsedInternalKey(
|
|
|
|
saved_key_.GetUserKey(), kMaxSequenceNumber, kValueTypeForSeek));
|
|
|
|
iter_->Seek(last_key.GetInternalKey());
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
direction_ = kForward;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Skip keys less than the current key() (a.k.a. saved_key_).
|
|
|
|
while (iter_->Valid()) {
|
|
|
|
ParsedInternalKey ikey;
|
|
|
|
if (!ParseKey(&ikey)) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (user_comparator_.Compare(ikey.user_key, saved_key_.GetUserKey()) >= 0) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
|
|
|
}
|
|
|
|
iter_->Next();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!iter_->status().ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
return false;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Move iter_ to the key before saved_key_.
|
|
|
|
bool DBIter::ReverseToBackward() {
|
|
|
|
assert(iter_->status().ok());
|
|
|
|
|
|
|
|
// When current_entry_is_merged_ is true, iter_ may be positioned on the next
|
|
|
|
// key, which may not exist or may have prefix different from current.
|
|
|
|
// If that's the case, seek to saved_key_.
|
|
|
|
if (current_entry_is_merged_ &&
|
|
|
|
((prefix_extractor_ != nullptr && !total_order_seek_) ||
|
|
|
|
!iter_->Valid())) {
|
|
|
|
IterKey last_key;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Using kMaxSequenceNumber and kValueTypeForSeek
|
|
|
|
// (not kValueTypeForSeekForPrev) to seek to a key strictly smaller
|
|
|
|
// than saved_key_.
|
|
|
|
last_key.SetInternalKey(ParsedInternalKey(
|
|
|
|
saved_key_.GetUserKey(), kMaxSequenceNumber, kValueTypeForSeek));
|
|
|
|
if (prefix_extractor_ != nullptr && !total_order_seek_) {
|
|
|
|
iter_->SeekForPrev(last_key.GetInternalKey());
|
|
|
|
} else {
|
|
|
|
// Some iterators may not support SeekForPrev(), so we avoid using it
|
|
|
|
// when prefix seek mode is disabled. This is somewhat expensive
|
|
|
|
// (an extra Prev(), as well as an extra change of direction of iter_),
|
|
|
|
// so we may need to reconsider it later.
|
|
|
|
iter_->Seek(last_key.GetInternalKey());
|
|
|
|
if (!iter_->Valid() && iter_->status().ok()) {
|
|
|
|
iter_->SeekToLast();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
direction_ = kReverse;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return FindUserKeyBeforeSavedKey();
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::PrevInternal() {
|
|
|
|
while (iter_->Valid()) {
|
|
|
|
saved_key_.SetUserKey(
|
|
|
|
ExtractUserKey(iter_->key()),
|
|
|
|
!iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
|
|
|
|
|
|
|
|
if (prefix_extractor_ && prefix_same_as_start_ &&
|
|
|
|
prefix_extractor_->Transform(saved_key_.GetUserKey())
|
|
|
|
.compare(prefix_start_key_) != 0) {
|
|
|
|
// Current key does not have the same prefix as start
|
|
|
|
valid_ = false;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (iterate_lower_bound_ != nullptr &&
|
|
|
|
user_comparator_.Compare(saved_key_.GetUserKey(),
|
|
|
|
*iterate_lower_bound_) < 0) {
|
|
|
|
// We've iterated earlier than the user-specified lower bound.
|
|
|
|
valid_ = false;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!FindValueForCurrentKey()) { // assigns valid_
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Whether or not we found a value for current key, we need iter_ to end up
|
|
|
|
// on a smaller key.
|
|
|
|
if (!FindUserKeyBeforeSavedKey()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (valid_) {
|
|
|
|
// Found the value.
|
|
|
|
return;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
if (TooManyInternalKeysSkipped(false)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
// We haven't found any key - iterator is not valid
|
|
|
|
valid_ = false;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Used for backwards iteration.
|
|
|
|
// Looks at the entries with user key saved_key_ and finds the most up-to-date
|
|
|
|
// value for it, or executes a merge, or determines that the value was deleted.
|
|
|
|
// Sets valid_ to true if the value is found and is ready to be presented to
|
|
|
|
// the user through value().
|
|
|
|
// Sets valid_ to false if the value was deleted, and we should try another key.
|
|
|
|
// Returns false if an error occurred, and !status().ok() and !valid_.
|
|
|
|
//
|
|
|
|
// PRE: iter_ is positioned on the last entry with user key equal to saved_key_.
|
|
|
|
// POST: iter_ is positioned on one of the entries equal to saved_key_, or on
|
|
|
|
// the entry just before them, or on the entry just after them.
|
|
|
|
bool DBIter::FindValueForCurrentKey() {
|
|
|
|
assert(iter_->Valid());
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
merge_context_.Clear();
|
|
|
|
current_entry_is_merged_ = false;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
// last entry before merge (could be kTypeDeletion, kTypeSingleDeletion or
|
|
|
|
// kTypeValue)
|
|
|
|
ValueType last_not_merge_type = kTypeDeletion;
|
|
|
|
ValueType last_key_entry_type = kTypeDeletion;
|
|
|
|
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
// Temporarily pin blocks that hold (merge operands / the value)
|
|
|
|
ReleaseTempPinnedData();
|
|
|
|
TempPinData();
|
|
|
|
size_t num_skipped = 0;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
while (iter_->Valid()) {
|
|
|
|
ParsedInternalKey ikey;
|
|
|
|
if (!ParseKey(&ikey)) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!IsVisible(ikey.sequence) ||
|
|
|
|
!user_comparator_.Equal(ikey.user_key, saved_key_.GetUserKey())) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (TooManyInternalKeysSkipped()) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// This user key has lots of entries.
|
|
|
|
// We're going from old to new, and it's taking too long. Let's do a Seek()
|
|
|
|
// and go from new to old. This helps when a key was overwritten many times.
|
|
|
|
if (num_skipped >= max_skip_ && CanReseekToSkip()) {
|
|
|
|
return FindValueForCurrentKeyUsingSeek();
|
|
|
|
}
|
|
|
|
|
|
|
|
last_key_entry_type = ikey.type;
|
|
|
|
switch (last_key_entry_type) {
|
|
|
|
case kTypeValue:
|
|
|
|
case kTypeBlobIndex:
|
Maintain position in range deletions map
Summary:
When deletion-collapsing mode is enabled (i.e., for DBIter/CompactionIterator), we maintain position in the tombstone maps across calls to ShouldDelete(). Since iterators often access keys sequentially (or reverse-sequentially), scanning forward/backward from the last position can be faster than binary-searching the map for every key.
- When Next() is invoked on an iterator, we use kForwardTraversal to scan forwards, if needed, until arriving at the range deletion containing the next key.
- Similarly for Prev(), we use kBackwardTraversal to scan backwards in the range deletion map.
- When the iterator seeks, we use kBinarySearch for repositioning
- After tombstones are added or before the first ShouldDelete() invocation, the current position is set to invalid, which forces kBinarySearch to be used.
- Non-iterator users (i.e., Get()) use kFullScan, which has the same behavior as before---scan the whole map for every key passed to ShouldDelete().
Closes https://github.com/facebook/rocksdb/pull/1701
Differential Revision: D4350318
Pulled By: ajkr
fbshipit-source-id: 5129b76
8 years ago
|
|
|
if (range_del_agg_.ShouldDelete(
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
ikey, RangeDelPositioningMode::kBackwardTraversal)) {
|
|
|
|
last_key_entry_type = kTypeRangeDeletion;
|
|
|
|
PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
|
|
|
|
} else {
|
|
|
|
assert(iter_->IsValuePinned());
|
|
|
|
pinned_value_ = iter_->value();
|
|
|
|
}
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
merge_context_.Clear();
|
|
|
|
last_not_merge_type = last_key_entry_type;
|
|
|
|
break;
|
|
|
|
case kTypeDeletion:
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
case kTypeSingleDeletion:
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
merge_context_.Clear();
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
last_not_merge_type = last_key_entry_type;
|
|
|
|
PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
|
|
|
|
break;
|
|
|
|
case kTypeMerge:
|
Maintain position in range deletions map
Summary:
When deletion-collapsing mode is enabled (i.e., for DBIter/CompactionIterator), we maintain position in the tombstone maps across calls to ShouldDelete(). Since iterators often access keys sequentially (or reverse-sequentially), scanning forward/backward from the last position can be faster than binary-searching the map for every key.
- When Next() is invoked on an iterator, we use kForwardTraversal to scan forwards, if needed, until arriving at the range deletion containing the next key.
- Similarly for Prev(), we use kBackwardTraversal to scan backwards in the range deletion map.
- When the iterator seeks, we use kBinarySearch for repositioning
- After tombstones are added or before the first ShouldDelete() invocation, the current position is set to invalid, which forces kBinarySearch to be used.
- Non-iterator users (i.e., Get()) use kFullScan, which has the same behavior as before---scan the whole map for every key passed to ShouldDelete().
Closes https://github.com/facebook/rocksdb/pull/1701
Differential Revision: D4350318
Pulled By: ajkr
fbshipit-source-id: 5129b76
8 years ago
|
|
|
if (range_del_agg_.ShouldDelete(
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
ikey, RangeDelPositioningMode::kBackwardTraversal)) {
|
|
|
|
merge_context_.Clear();
|
|
|
|
last_key_entry_type = kTypeRangeDeletion;
|
|
|
|
last_not_merge_type = last_key_entry_type;
|
|
|
|
PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
|
|
|
|
} else {
|
|
|
|
assert(merge_operator_ != nullptr);
|
|
|
|
merge_context_.PushOperandBack(
|
|
|
|
iter_->value(), iter_->IsValuePinned() /* operand_pinned */);
|
|
|
|
PERF_COUNTER_ADD(internal_merge_count, 1);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
assert(false);
|
|
|
|
}
|
|
|
|
|
|
|
|
PERF_COUNTER_ADD(internal_key_skipped_count, 1);
|
|
|
|
iter_->Prev();
|
|
|
|
++num_skipped;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
}
|
|
|
|
|
|
|
|
if (!iter_->status().ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status s;
|
|
|
|
is_blob_ = false;
|
|
|
|
switch (last_key_entry_type) {
|
|
|
|
case kTypeDeletion:
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
9 years ago
|
|
|
case kTypeSingleDeletion:
|
|
|
|
case kTypeRangeDeletion:
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
|
|
|
case kTypeMerge:
|
|
|
|
current_entry_is_merged_ = true;
|
|
|
|
if (last_not_merge_type == kTypeDeletion ||
|
|
|
|
last_not_merge_type == kTypeSingleDeletion ||
|
|
|
|
last_not_merge_type == kTypeRangeDeletion) {
|
|
|
|
s = MergeHelper::TimedFullMerge(
|
|
|
|
merge_operator_, saved_key_.GetUserKey(), nullptr,
|
|
|
|
merge_context_.GetOperands(), &saved_value_, logger_, statistics_,
|
|
|
|
env_, &pinned_value_, true);
|
|
|
|
} else if (last_not_merge_type == kTypeBlobIndex) {
|
|
|
|
if (!allow_blob_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
|
|
|
|
status_ = Status::NotSupported(
|
|
|
|
"Encounter unexpected blob index. Please open DB with "
|
|
|
|
"rocksdb::blob_db::BlobDB instead.");
|
|
|
|
} else {
|
|
|
|
status_ =
|
|
|
|
Status::NotSupported("Blob DB does not support merge operator.");
|
|
|
|
}
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
assert(last_not_merge_type == kTypeValue);
|
|
|
|
s = MergeHelper::TimedFullMerge(
|
|
|
|
merge_operator_, saved_key_.GetUserKey(), &pinned_value_,
|
|
|
|
merge_context_.GetOperands(), &saved_value_, logger_, statistics_,
|
|
|
|
env_, &pinned_value_, true);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case kTypeValue:
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// do nothing - we've already has value in pinned_value_
|
|
|
|
break;
|
|
|
|
case kTypeBlobIndex:
|
|
|
|
if (!allow_blob_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
|
|
|
|
status_ = Status::NotSupported(
|
|
|
|
"Encounter unexpected blob index. Please open DB with "
|
|
|
|
"rocksdb::blob_db::BlobDB instead.");
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
is_blob_ = true;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
assert(false);
|
|
|
|
break;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!s.ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
status_ = s;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
valid_ = true;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
// This function is used in FindValueForCurrentKey.
|
|
|
|
// We use Seek() function instead of Prev() to find necessary value
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// TODO: This is very similar to FindNextUserEntry() and MergeValuesNewToOld().
|
|
|
|
// Would be nice to reuse some code.
|
|
|
|
bool DBIter::FindValueForCurrentKeyUsingSeek() {
|
Introduce FullMergeV2 (eliminate memcpy from merge operators)
Summary:
This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice>
This diff is stacked on top of D56493 and D56511
In this diff we
- Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future
- Replace std::deque<std::string> with std::vector<Slice> to pass operands
- Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187)
- Allow FullMergeV2 output to be an existing operand
```
[Everything in Memtable | 10K operands | 10 KB each | 1 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s
readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s
readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s
readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s
readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s
[master]
readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s
readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s
readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s
readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s
readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s
```
```
[Everything in Memtable | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000
[FullMergeV2]
readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s
readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s
readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s
readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s
readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s
[master]
readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s
readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s
readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s
readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s
readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 1 operand per key]
[FullMergeV2]
$ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s
readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s
readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s
readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s
readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s
[master]
readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s
readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s
readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s
readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s
readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s
```
```
[Everything in Block cache | 10K operands | 10 KB each | 10 operand per key]
DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions
[FullMergeV2]
readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s
readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s
readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s
readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s
readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s
[master]
readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s
readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s
readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s
readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s
readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s
```
Test Plan: COMPILE_WITH_ASAN=1 make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: lovro, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57075
8 years ago
|
|
|
// FindValueForCurrentKey will enable pinning before calling
|
|
|
|
// FindValueForCurrentKeyUsingSeek()
|
|
|
|
assert(pinned_iters_mgr_.PinningEnabled());
|
|
|
|
std::string last_key;
|
|
|
|
AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetUserKey(),
|
|
|
|
sequence_, kValueTypeForSeek));
|
|
|
|
iter_->Seek(last_key);
|
|
|
|
RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// In case read_callback presents, the value we seek to may not be visible.
|
|
|
|
// Find the next value that's visible.
|
|
|
|
ParsedInternalKey ikey;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
while (true) {
|
|
|
|
if (!iter_->Valid()) {
|
|
|
|
valid_ = false;
|
|
|
|
return iter_->status().ok();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!ParseKey(&ikey)) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (!user_comparator_.Equal(ikey.user_key, saved_key_.GetUserKey())) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// No visible values for this key, even though FindValueForCurrentKey()
|
|
|
|
// has seen some. This is possible if we're using a tailing iterator, and
|
|
|
|
// the entries were discarded in a compaction.
|
|
|
|
valid_ = false;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (IsVisible(ikey.sequence)) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
iter_->Next();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion ||
|
Maintain position in range deletions map
Summary:
When deletion-collapsing mode is enabled (i.e., for DBIter/CompactionIterator), we maintain position in the tombstone maps across calls to ShouldDelete(). Since iterators often access keys sequentially (or reverse-sequentially), scanning forward/backward from the last position can be faster than binary-searching the map for every key.
- When Next() is invoked on an iterator, we use kForwardTraversal to scan forwards, if needed, until arriving at the range deletion containing the next key.
- Similarly for Prev(), we use kBackwardTraversal to scan backwards in the range deletion map.
- When the iterator seeks, we use kBinarySearch for repositioning
- After tombstones are added or before the first ShouldDelete() invocation, the current position is set to invalid, which forces kBinarySearch to be used.
- Non-iterator users (i.e., Get()) use kFullScan, which has the same behavior as before---scan the whole map for every key passed to ShouldDelete().
Closes https://github.com/facebook/rocksdb/pull/1701
Differential Revision: D4350318
Pulled By: ajkr
fbshipit-source-id: 5129b76
8 years ago
|
|
|
range_del_agg_.ShouldDelete(
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
ikey, RangeDelPositioningMode::kBackwardTraversal)) {
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
|
|
|
}
|
|
|
|
if (ikey.type == kTypeBlobIndex && !allow_blob_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
|
|
|
|
status_ = Status::NotSupported(
|
|
|
|
"Encounter unexpected blob index. Please open DB with "
|
|
|
|
"rocksdb::blob_db::BlobDB instead.");
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (ikey.type == kTypeValue || ikey.type == kTypeBlobIndex) {
|
|
|
|
assert(iter_->IsValuePinned());
|
|
|
|
pinned_value_ = iter_->value();
|
|
|
|
valid_ = true;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
// kTypeMerge. We need to collect all kTypeMerge values and save them
|
|
|
|
// in operands
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
assert(ikey.type == kTypeMerge);
|
|
|
|
current_entry_is_merged_ = true;
|
Eliminate std::deque initialization while iterating over merge operands
Summary:
This patch is similar to D52563, When we iterate over a DB with merge operands we keep creating std::queue to store the operands, optimize this by reusing merge_operands_ data member
Before the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.757 micros/op 266141 ops/sec; 29.4 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.413 micros/op 2423538 ops/sec; 268.1 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.451 micros/op 2219071 ops/sec; 245.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.420 micros/op 2382039 ops/sec; 263.5 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.408 micros/op 2452017 ops/sec; 271.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.947 micros/op 253376 ops/sec; 28.0 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.441 micros/op 2266473 ops/sec; 250.7 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.471 micros/op 2122033 ops/sec; 234.8 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.440 micros/op 2271407 ops/sec; 251.3 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.429 micros/op 2331471 ops/sec; 257.9 MB/s
```
with the patch
```
./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq" --db="/dev/shm/bench_merge_memcpy_on_the_fly/" --merge_operator="put" --merge_keys=10000 --num=10000
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 4.080 micros/op 245092 ops/sec; 27.1 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.308 micros/op 3241843 ops/sec; 358.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.312 micros/op 3200408 ops/sec; 354.0 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.332 micros/op 3013962 ops/sec; 333.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.300 micros/op 3328017 ops/sec; 368.2 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
mergerandom : 3.973 micros/op 251705 ops/sec; 27.8 MB/s ( updates:10000)
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.320 micros/op 3123752 ops/sec; 345.6 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.335 micros/op 2986641 ops/sec; 330.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.339 micros/op 2950047 ops/sec; 326.4 MB/s
DB path: [/dev/shm/bench_merge_memcpy_on_the_fly/]
readseq : 0.319 micros/op 3131565 ops/sec; 346.4 MB/s
```
Test Plan: make check -j64
Reviewers: yhchiang, andrewkr, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56031
9 years ago
|
|
|
merge_context_.Clear();
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
merge_context_.PushOperand(iter_->value(),
|
|
|
|
iter_->IsValuePinned() /* operand_pinned */);
|
|
|
|
while (true) {
|
|
|
|
iter_->Next();
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!iter_->Valid()) {
|
|
|
|
if (!iter_->status().ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (!ParseKey(&ikey)) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (!user_comparator_.Equal(ikey.user_key, saved_key_.GetUserKey())) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion ||
|
|
|
|
range_del_agg_.ShouldDelete(
|
|
|
|
ikey, RangeDelPositioningMode::kForwardTraversal)) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
break;
|
|
|
|
} else if (ikey.type == kTypeValue) {
|
|
|
|
const Slice val = iter_->value();
|
|
|
|
Status s = MergeHelper::TimedFullMerge(
|
|
|
|
merge_operator_, saved_key_.GetUserKey(), &val,
|
|
|
|
merge_context_.GetOperands(), &saved_value_, logger_, statistics_,
|
|
|
|
env_, &pinned_value_, true);
|
|
|
|
if (!s.ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
status_ = s;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
valid_ = true;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
|
|
|
} else if (ikey.type == kTypeMerge) {
|
|
|
|
merge_context_.PushOperand(iter_->value(),
|
|
|
|
iter_->IsValuePinned() /* operand_pinned */);
|
|
|
|
PERF_COUNTER_ADD(internal_merge_count, 1);
|
|
|
|
} else if (ikey.type == kTypeBlobIndex) {
|
|
|
|
if (!allow_blob_) {
|
|
|
|
ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
|
|
|
|
status_ = Status::NotSupported(
|
|
|
|
"Encounter unexpected blob index. Please open DB with "
|
|
|
|
"rocksdb::blob_db::BlobDB instead.");
|
|
|
|
} else {
|
|
|
|
status_ =
|
|
|
|
Status::NotSupported("Blob DB does not support merge operator.");
|
|
|
|
}
|
|
|
|
valid_ = false;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
assert(false);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
Status s = MergeHelper::TimedFullMerge(
|
|
|
|
merge_operator_, saved_key_.GetUserKey(), nullptr,
|
|
|
|
merge_context_.GetOperands(), &saved_value_, logger_, statistics_, env_,
|
|
|
|
&pinned_value_, true);
|
|
|
|
if (!s.ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
status_ = s;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Make sure we leave iter_ in a good state. If it's valid and we don't care
|
|
|
|
// about prefixes, that's already good enough. Otherwise it needs to be
|
|
|
|
// seeked to the current key.
|
|
|
|
if ((prefix_extractor_ != nullptr && !total_order_seek_) || !iter_->Valid()) {
|
|
|
|
if (prefix_extractor_ != nullptr && !total_order_seek_) {
|
|
|
|
iter_->SeekForPrev(last_key);
|
|
|
|
} else {
|
|
|
|
iter_->Seek(last_key);
|
|
|
|
if (!iter_->Valid() && iter_->status().ok()) {
|
|
|
|
iter_->SeekToLast();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
valid_ = true;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Move backwards until the key smaller than saved_key_.
|
|
|
|
// Changes valid_ only if return value is false.
|
|
|
|
bool DBIter::FindUserKeyBeforeSavedKey() {
|
|
|
|
assert(status_.ok());
|
|
|
|
size_t num_skipped = 0;
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
while (iter_->Valid()) {
|
|
|
|
ParsedInternalKey ikey;
|
|
|
|
if (!ParseKey(&ikey)) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (user_comparator_.Compare(ikey.user_key, saved_key_.GetUserKey()) < 0) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (TooManyInternalKeysSkipped()) {
|
|
|
|
return false;
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
assert(ikey.sequence != kMaxSequenceNumber);
|
|
|
|
if (!IsVisible(ikey.sequence)) {
|
|
|
|
PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
|
|
|
|
} else {
|
|
|
|
PERF_COUNTER_ADD(internal_key_skipped_count, 1);
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
if (num_skipped >= max_skip_ && CanReseekToSkip()) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
num_skipped = 0;
|
|
|
|
IterKey last_key;
|
|
|
|
last_key.SetInternalKey(ParsedInternalKey(
|
|
|
|
saved_key_.GetUserKey(), kMaxSequenceNumber, kValueTypeForSeek));
|
|
|
|
// It would be more efficient to use SeekForPrev() here, but some
|
|
|
|
// iterators may not support it.
|
|
|
|
iter_->Seek(last_key.GetInternalKey());
|
|
|
|
RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
|
|
|
|
if (!iter_->Valid()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
++num_skipped;
|
|
|
|
}
|
|
|
|
|
|
|
|
iter_->Prev();
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
|
|
|
|
if (!iter_->status().ok()) {
|
|
|
|
valid_ = false;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool DBIter::TooManyInternalKeysSkipped(bool increment) {
|
|
|
|
if ((max_skippable_internal_keys_ > 0) &&
|
|
|
|
(num_internal_keys_skipped_ > max_skippable_internal_keys_)) {
|
|
|
|
valid_ = false;
|
|
|
|
status_ = Status::Incomplete("Too many internal keys skipped.");
|
|
|
|
return true;
|
|
|
|
} else if (increment) {
|
|
|
|
num_internal_keys_skipped_++;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool DBIter::IsVisible(SequenceNumber sequence) {
|
|
|
|
if (read_callback_ == nullptr) {
|
|
|
|
return sequence <= sequence_;
|
|
|
|
} else {
|
|
|
|
return read_callback_->IsVisible(sequence);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
bool DBIter::CanReseekToSkip() {
|
|
|
|
return read_callback_ == nullptr || read_callback_->CanReseekToSkip();
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::Seek(const Slice& target) {
|
|
|
|
PERF_CPU_TIMER_GUARD(iter_seek_cpu_nanos, env_);
|
|
|
|
StopWatch sw(env_, statistics_, DB_SEEK);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
status_ = Status::OK();
|
|
|
|
ReleaseTempPinnedData();
|
|
|
|
ResetInternalKeysSkippedCounter();
|
|
|
|
|
|
|
|
SequenceNumber seq = sequence_;
|
|
|
|
saved_key_.Clear();
|
|
|
|
saved_key_.SetInternalKey(target, seq);
|
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
if (db_impl_ != nullptr && cfd_ != nullptr) {
|
|
|
|
db_impl_->TraceIteratorSeek(cfd_->GetID(), target);
|
|
|
|
}
|
|
|
|
#endif // ROCKSDB_LITE
|
|
|
|
|
|
|
|
if (iterate_lower_bound_ != nullptr &&
|
|
|
|
user_comparator_.Compare(saved_key_.GetUserKey(), *iterate_lower_bound_) <
|
|
|
|
0) {
|
|
|
|
saved_key_.Clear();
|
|
|
|
saved_key_.SetInternalKey(*iterate_lower_bound_, seq);
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
PERF_TIMER_GUARD(seek_internal_seek_time);
|
|
|
|
iter_->Seek(saved_key_.GetInternalKey());
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
range_del_agg_.InvalidateRangeDelMapPositions();
|
|
|
|
}
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK);
|
|
|
|
if (iter_->Valid()) {
|
|
|
|
if (prefix_extractor_ && prefix_same_as_start_) {
|
|
|
|
prefix_start_key_ = prefix_extractor_->Transform(target);
|
|
|
|
}
|
|
|
|
direction_ = kForward;
|
|
|
|
ClearSavedValue();
|
|
|
|
FindNextUserEntry(false /* not skipping */, prefix_same_as_start_);
|
|
|
|
if (!valid_) {
|
|
|
|
prefix_start_key_.clear();
|
|
|
|
}
|
|
|
|
if (statistics_ != nullptr) {
|
|
|
|
if (valid_) {
|
|
|
|
// Decrement since we don't want to count this key as skipped
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
|
|
|
|
RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
|
|
|
|
PERF_COUNTER_ADD(iter_read_bytes, key().size() + value().size());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
valid_ = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
|
|
|
|
prefix_start_buf_.SetUserKey(prefix_start_key_);
|
|
|
|
prefix_start_key_ = prefix_start_buf_.GetUserKey();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::SeekForPrev(const Slice& target) {
|
|
|
|
PERF_CPU_TIMER_GUARD(iter_seek_cpu_nanos, env_);
|
|
|
|
StopWatch sw(env_, statistics_, DB_SEEK);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
status_ = Status::OK();
|
|
|
|
ReleaseTempPinnedData();
|
|
|
|
ResetInternalKeysSkippedCounter();
|
|
|
|
saved_key_.Clear();
|
|
|
|
// now saved_key is used to store internal key.
|
|
|
|
saved_key_.SetInternalKey(target, 0 /* sequence_number */,
|
|
|
|
kValueTypeForSeekForPrev);
|
|
|
|
|
|
|
|
if (iterate_upper_bound_ != nullptr &&
|
|
|
|
user_comparator_.Compare(saved_key_.GetUserKey(),
|
|
|
|
*iterate_upper_bound_) >= 0) {
|
|
|
|
saved_key_.Clear();
|
|
|
|
saved_key_.SetInternalKey(*iterate_upper_bound_, kMaxSequenceNumber);
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
PERF_TIMER_GUARD(seek_internal_seek_time);
|
|
|
|
iter_->SeekForPrev(saved_key_.GetInternalKey());
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
range_del_agg_.InvalidateRangeDelMapPositions();
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
if (db_impl_ != nullptr && cfd_ != nullptr) {
|
|
|
|
db_impl_->TraceIteratorSeekForPrev(cfd_->GetID(), target);
|
|
|
|
}
|
|
|
|
#endif // ROCKSDB_LITE
|
|
|
|
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK);
|
|
|
|
if (iter_->Valid()) {
|
|
|
|
if (prefix_extractor_ && prefix_same_as_start_) {
|
|
|
|
prefix_start_key_ = prefix_extractor_->Transform(target);
|
|
|
|
}
|
|
|
|
direction_ = kReverse;
|
|
|
|
ClearSavedValue();
|
|
|
|
PrevInternal();
|
|
|
|
if (!valid_) {
|
|
|
|
prefix_start_key_.clear();
|
|
|
|
}
|
|
|
|
if (statistics_ != nullptr) {
|
|
|
|
if (valid_) {
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
|
|
|
|
RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
|
|
|
|
PERF_COUNTER_ADD(iter_read_bytes, key().size() + value().size());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
valid_ = false;
|
|
|
|
}
|
|
|
|
if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
|
|
|
|
prefix_start_buf_.SetUserKey(prefix_start_key_);
|
|
|
|
prefix_start_key_ = prefix_start_buf_.GetUserKey();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::SeekToFirst() {
|
|
|
|
if (iterate_lower_bound_ != nullptr) {
|
|
|
|
Seek(*iterate_lower_bound_);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
PERF_CPU_TIMER_GUARD(iter_seek_cpu_nanos, env_);
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
// Don't use iter_::Seek() if we set a prefix extractor
|
|
|
|
// because prefix seek will be used.
|
|
|
|
if (prefix_extractor_ != nullptr && !total_order_seek_) {
|
|
|
|
max_skip_ = std::numeric_limits<uint64_t>::max();
|
|
|
|
}
|
|
|
|
status_ = Status::OK();
|
|
|
|
direction_ = kForward;
|
|
|
|
ReleaseTempPinnedData();
|
|
|
|
ResetInternalKeysSkippedCounter();
|
|
|
|
ClearSavedValue();
|
|
|
|
|
|
|
|
{
|
|
|
|
PERF_TIMER_GUARD(seek_internal_seek_time);
|
|
|
|
iter_->SeekToFirst();
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
range_del_agg_.InvalidateRangeDelMapPositions();
|
|
|
|
}
|
|
|
|
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK);
|
|
|
|
if (iter_->Valid()) {
|
|
|
|
saved_key_.SetUserKey(
|
|
|
|
ExtractUserKey(iter_->key()),
|
|
|
|
!iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
|
|
|
|
FindNextUserEntry(false /* not skipping */, false /* no prefix check */);
|
|
|
|
if (statistics_ != nullptr) {
|
|
|
|
if (valid_) {
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
|
|
|
|
RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
|
|
|
|
PERF_COUNTER_ADD(iter_read_bytes, key().size() + value().size());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
valid_ = false;
|
|
|
|
}
|
|
|
|
if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
|
|
|
|
prefix_start_buf_.SetUserKey(
|
|
|
|
prefix_extractor_->Transform(saved_key_.GetUserKey()));
|
|
|
|
prefix_start_key_ = prefix_start_buf_.GetUserKey();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBIter::SeekToLast() {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (iterate_upper_bound_ != nullptr) {
|
|
|
|
// Seek to last key strictly less than ReadOptions.iterate_upper_bound.
|
|
|
|
SeekForPrev(*iterate_upper_bound_);
|
|
|
|
if (Valid() && user_comparator_.Equal(*iterate_upper_bound_, key())) {
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
ReleaseTempPinnedData();
|
|
|
|
PrevInternal();
|
|
|
|
}
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
PERF_CPU_TIMER_GUARD(iter_seek_cpu_nanos, env_);
|
|
|
|
// Don't use iter_::Seek() if we set a prefix extractor
|
|
|
|
// because prefix seek will be used.
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
if (prefix_extractor_ != nullptr && !total_order_seek_) {
|
|
|
|
max_skip_ = std::numeric_limits<uint64_t>::max();
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
status_ = Status::OK();
|
|
|
|
direction_ = kReverse;
|
|
|
|
ReleaseTempPinnedData();
|
|
|
|
ResetInternalKeysSkippedCounter();
|
|
|
|
ClearSavedValue();
|
|
|
|
|
|
|
|
{
|
|
|
|
PERF_TIMER_GUARD(seek_internal_seek_time);
|
|
|
|
iter_->SeekToLast();
|
Range deletion performance improvements + cleanup (#4014)
Summary:
This fixes the same performance issue that #3992 fixes but with much more invasive cleanup.
I'm more excited about this PR because it paves the way for fixing another problem we uncovered at Cockroach where range deletion tombstones can cause massive compactions. For example, suppose L4 contains deletions from [a, c) and [x, z) and no other keys, and L5 is entirely empty. L6, however, is full of data. When compacting L4 -> L5, we'll end up with one file that spans, massively, from [a, z). When we go to compact L5 -> L6, we'll have to rewrite all of L6! If, instead of range deletions in L4, we had keys a, b, x, y, and z, RocksDB would have been smart enough to create two files in L5: one for a and b and another for x, y, and z.
With the changes in this PR, it will be possible to adjust the compaction logic to split tombstones/start new output files when they would span too many files in the grandparent level.
ajkr please take a look when you have a minute!
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4014
Differential Revision: D8773253
Pulled By: ajkr
fbshipit-source-id: ec62fa85f648fdebe1380b83ed997f9baec35677
6 years ago
|
|
|
range_del_agg_.InvalidateRangeDelMapPositions();
|
|
|
|
}
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
7 years ago
|
|
|
PrevInternal();
|
|
|
|
if (statistics_ != nullptr) {
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK);
|
|
|
|
if (valid_) {
|
|
|
|
RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
|
|
|
|
RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
|
|
|
|
PERF_COUNTER_ADD(iter_read_bytes, key().size() + value().size());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
|
|
|
|
prefix_start_buf_.SetUserKey(
|
|
|
|
prefix_extractor_->Transform(saved_key_.GetUserKey()));
|
|
|
|
prefix_start_key_ = prefix_start_buf_.GetUserKey();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Iterator* NewDBIterator(Env* env, const ReadOptions& read_options,
|
|
|
|
const ImmutableCFOptions& cf_options,
|
|
|
|
const MutableCFOptions& mutable_cf_options,
|
|
|
|
const Comparator* user_key_comparator,
|
|
|
|
InternalIterator* internal_iter,
|
|
|
|
const SequenceNumber& sequence,
|
|
|
|
uint64_t max_sequential_skip_in_iterations,
|
|
|
|
ReadCallback* read_callback, DBImpl* db_impl,
|
|
|
|
ColumnFamilyData* cfd, bool allow_blob) {
|
|
|
|
DBIter* db_iter = new DBIter(
|
|
|
|
env, read_options, cf_options, mutable_cf_options, user_key_comparator,
|
|
|
|
internal_iter, sequence, false, max_sequential_skip_in_iterations,
|
|
|
|
read_callback, db_impl, cfd, allow_blob);
|
|
|
|
return db_iter;
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
}
|
|
|
|
|
|
|
|
ArenaWrappedDBIter::~ArenaWrappedDBIter() { db_iter_->~DBIter(); }
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
|
|
|
|
ReadRangeDelAggregator* ArenaWrappedDBIter::GetRangeDelAggregator() {
|
|
|
|
return db_iter_->GetRangeDelAggregator();
|
|
|
|
}
|
|
|
|
|
|
|
|
void ArenaWrappedDBIter::SetIterUnderDBIter(InternalIterator* iter) {
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
static_cast<DBIter*>(db_iter_)->SetIter(iter);
|
|
|
|
}
|
|
|
|
|
|
|
|
inline bool ArenaWrappedDBIter::Valid() const { return db_iter_->Valid(); }
|
|
|
|
inline void ArenaWrappedDBIter::SeekToFirst() { db_iter_->SeekToFirst(); }
|
|
|
|
inline void ArenaWrappedDBIter::SeekToLast() { db_iter_->SeekToLast(); }
|
|
|
|
inline void ArenaWrappedDBIter::Seek(const Slice& target) {
|
|
|
|
db_iter_->Seek(target);
|
|
|
|
}
|
|
|
|
inline void ArenaWrappedDBIter::SeekForPrev(const Slice& target) {
|
|
|
|
db_iter_->SeekForPrev(target);
|
|
|
|
}
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
inline void ArenaWrappedDBIter::Next() { db_iter_->Next(); }
|
|
|
|
inline void ArenaWrappedDBIter::Prev() { db_iter_->Prev(); }
|
|
|
|
inline Slice ArenaWrappedDBIter::key() const { return db_iter_->key(); }
|
|
|
|
inline Slice ArenaWrappedDBIter::value() const { return db_iter_->value(); }
|
|
|
|
inline Status ArenaWrappedDBIter::status() const { return db_iter_->status(); }
|
|
|
|
bool ArenaWrappedDBIter::IsBlob() const { return db_iter_->IsBlob(); }
|
|
|
|
inline Status ArenaWrappedDBIter::GetProperty(std::string prop_name,
|
|
|
|
std::string* prop) {
|
|
|
|
if (prop_name == "rocksdb.iterator.super-version-number") {
|
|
|
|
// First try to pass the value returned from inner iterator.
|
|
|
|
if (!db_iter_->GetProperty(prop_name, prop).ok()) {
|
|
|
|
*prop = ToString(sv_number_);
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
return db_iter_->GetProperty(prop_name, prop);
|
|
|
|
}
|
|
|
|
|
|
|
|
void ArenaWrappedDBIter::Init(Env* env, const ReadOptions& read_options,
|
|
|
|
const ImmutableCFOptions& cf_options,
|
|
|
|
const MutableCFOptions& mutable_cf_options,
|
|
|
|
const SequenceNumber& sequence,
|
|
|
|
uint64_t max_sequential_skip_in_iteration,
|
|
|
|
uint64_t version_number,
|
|
|
|
ReadCallback* read_callback, DBImpl* db_impl,
|
|
|
|
ColumnFamilyData* cfd, bool allow_blob,
|
|
|
|
bool allow_refresh) {
|
|
|
|
auto mem = arena_.AllocateAligned(sizeof(DBIter));
|
|
|
|
db_iter_ = new (mem) DBIter(env, read_options, cf_options, mutable_cf_options,
|
|
|
|
cf_options.user_comparator, nullptr, sequence,
|
|
|
|
true, max_sequential_skip_in_iteration,
|
|
|
|
read_callback, db_impl, cfd, allow_blob);
|
|
|
|
sv_number_ = version_number;
|
|
|
|
allow_refresh_ = allow_refresh;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status ArenaWrappedDBIter::Refresh() {
|
|
|
|
if (cfd_ == nullptr || db_impl_ == nullptr || !allow_refresh_) {
|
|
|
|
return Status::NotSupported("Creating renew iterator is not allowed.");
|
|
|
|
}
|
|
|
|
assert(db_iter_ != nullptr);
|
|
|
|
// TODO(yiwu): For last_seq_same_as_publish_seq_==false, this is not the
|
|
|
|
// correct behavior. Will be corrected automatically when we take a snapshot
|
|
|
|
// here for the case of WritePreparedTxnDB.
|
|
|
|
SequenceNumber latest_seq = db_impl_->GetLatestSequenceNumber();
|
|
|
|
uint64_t cur_sv_number = cfd_->GetSuperVersionNumber();
|
|
|
|
if (sv_number_ != cur_sv_number) {
|
|
|
|
Env* env = db_iter_->env();
|
|
|
|
db_iter_->~DBIter();
|
|
|
|
arena_.~Arena();
|
|
|
|
new (&arena_) Arena();
|
|
|
|
|
|
|
|
SuperVersion* sv = cfd_->GetReferencedSuperVersion(db_impl_->mutex());
|
|
|
|
if (read_callback_) {
|
|
|
|
read_callback_->Refresh(latest_seq);
|
|
|
|
}
|
|
|
|
Init(env, read_options_, *(cfd_->ioptions()), sv->mutable_cf_options,
|
|
|
|
latest_seq, sv->mutable_cf_options.max_sequential_skip_in_iterations,
|
|
|
|
cur_sv_number, read_callback_, db_impl_, cfd_, allow_blob_,
|
|
|
|
allow_refresh_);
|
|
|
|
|
|
|
|
InternalIterator* internal_iter = db_impl_->NewInternalIterator(
|
|
|
|
read_options_, cfd_, sv, &arena_, db_iter_->GetRangeDelAggregator(),
|
|
|
|
latest_seq);
|
|
|
|
SetIterUnderDBIter(internal_iter);
|
|
|
|
} else {
|
|
|
|
db_iter_->set_sequence(latest_seq);
|
|
|
|
db_iter_->set_valid(false);
|
|
|
|
}
|
|
|
|
return Status::OK();
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
}
|
|
|
|
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
ArenaWrappedDBIter* NewArenaWrappedDbIterator(
|
|
|
|
Env* env, const ReadOptions& read_options,
|
|
|
|
const ImmutableCFOptions& cf_options,
|
|
|
|
const MutableCFOptions& mutable_cf_options, const SequenceNumber& sequence,
|
|
|
|
uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
|
|
|
|
ReadCallback* read_callback, DBImpl* db_impl, ColumnFamilyData* cfd,
|
|
|
|
bool allow_blob, bool allow_refresh) {
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
|
|
|
|
iter->Init(env, read_options, cf_options, mutable_cf_options, sequence,
|
|
|
|
max_sequential_skip_in_iterations, version_number, read_callback,
|
|
|
|
db_impl, cfd, allow_blob, allow_refresh);
|
|
|
|
if (db_impl != nullptr && cfd != nullptr && allow_refresh) {
|
|
|
|
iter->StoreRefreshInfo(read_options, db_impl, cfd, read_callback,
|
|
|
|
allow_blob);
|
|
|
|
}
|
|
|
|
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
11 years ago
|
|
|
return iter;
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace rocksdb
|