You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/db/db_with_timestamp_compactio...

180 lines
5.9 KiB

// 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/compaction/compaction.h"
#include "db/db_test_util.h"
#include "port/stack_trace.h"
#include "test_util/testutil.h"
namespace ROCKSDB_NAMESPACE {
namespace {
std::string Key1(uint64_t key) {
std::string ret;
PutFixed64(&ret, key);
std::reverse(ret.begin(), ret.end());
return ret;
}
std::string Timestamp(uint64_t ts) {
std::string ret;
PutFixed64(&ret, ts);
return ret;
}
} // anonymous namespace
class TimestampCompatibleCompactionTest : public DBTestBase {
public:
TimestampCompatibleCompactionTest()
: DBTestBase("ts_compatible_compaction_test", /*env_do_fsync=*/true) {}
std::string Get(const std::string& key, uint64_t ts) {
ReadOptions read_opts;
std::string ts_str = Timestamp(ts);
Slice ts_slice = ts_str;
read_opts.timestamp = &ts_slice;
std::string value;
Status s = db_->Get(read_opts, key, &value);
if (s.IsNotFound()) {
value.assign("NOT_FOUND");
} else if (!s.ok()) {
value.assign(s.ToString());
}
return value;
}
};
TEST_F(TimestampCompatibleCompactionTest, UserKeyCrossFileBoundary) {
Options options = CurrentOptions();
options.env = env_;
options.compaction_style = kCompactionStyleLevel;
options.comparator = test::BytewiseComparatorWithU64TsWrapper();
options.level0_file_num_compaction_trigger = 3;
constexpr size_t kNumKeysPerFile = 101;
options.memtable_factory.reset(
test::NewSpecialSkipListFactory(kNumKeysPerFile));
DestroyAndReopen(options);
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
SyncPoint::GetInstance()->SetCallBack(
"LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
const auto* compaction = reinterpret_cast<Compaction*>(arg);
ASSERT_NE(nullptr, compaction);
ASSERT_EQ(0, compaction->start_level());
ASSERT_EQ(1, compaction->num_input_levels());
// Check that all 3 L0 ssts are picked for level compaction.
ASSERT_EQ(3, compaction->num_input_files(0));
});
SyncPoint::GetInstance()->EnableProcessing();
// Write a L0 with keys 0, 1, ..., 99 with ts from 100 to 199.
uint64_t ts = 100;
uint64_t key = 0;
WriteOptions write_opts;
for (; key < kNumKeysPerFile - 1; ++key, ++ts) {
std::string ts_str = Timestamp(ts);
Revise APIs related to user-defined timestamp (#8946) Summary: ajkr reminded me that we have a rule of not including per-kv related data in `WriteOptions`. Namely, `WriteOptions` should not include information about "what-to-write", but should just include information about "how-to-write". According to this rule, `WriteOptions::timestamp` (experimental) is clearly a violation. Therefore, this PR removes `WriteOptions::timestamp` for compliance. After the removal, we need to pass timestamp info via another set of APIs. This PR proposes a set of overloaded functions `Put(write_opts, key, value, ts)`, `Delete(write_opts, key, ts)`, and `SingleDelete(write_opts, key, ts)`. Planned to add `Write(write_opts, batch, ts)`, but its complexity made me reconsider doing it in another PR (maybe). For better checking and returning error early, we also add a new set of APIs to `WriteBatch` that take extra `timestamp` information when writing to `WriteBatch`es. These set of APIs in `WriteBatchWithIndex` are currently not supported, and are on our TODO list. Removed `WriteBatch::AssignTimestamps()` and renamed `WriteBatch::AssignTimestamp()` to `WriteBatch::UpdateTimestamps()` since this method require that all keys have space for timestamps allocated already and multiple timestamps can be updated. The constructor of `WriteBatch` now takes a fourth argument `default_cf_ts_sz` which is the timestamp size of the default column family. This will be used to allocate space when calling APIs that do not specify a column family handle. Also, updated `DB::Get()`, `DB::MultiGet()`, `DB::NewIterator()`, `DB::NewIterators()` methods, replacing some assertions about timestamp to returning Status code. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8946 Test Plan: make check ./db_bench -benchmarks=fillseq,fillrandom,readrandom,readseq,deleterandom -user_timestamp_size=8 ./db_stress --user_timestamp_size=8 -nooverwritepercent=0 -test_secondary=0 -secondary_catch_up_one_in=0 -continuous_verification_interval=0 Make sure there is no perf regression by running the following ``` ./db_bench_opt -db=/dev/shm/rocksdb -use_existing_db=0 -level0_stop_writes_trigger=256 -level0_slowdown_writes_trigger=256 -level0_file_num_compaction_trigger=256 -disable_wal=1 -duration=10 -benchmarks=fillrandom ``` Before this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.831 micros/op 546235 ops/sec; 60.4 MB/s ``` After this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.820 micros/op 549404 ops/sec; 60.8 MB/s ``` Reviewed By: ltamasi Differential Revision: D33721359 Pulled By: riversand963 fbshipit-source-id: c131561534272c120ffb80711d42748d21badf09
3 years ago
ASSERT_OK(
db_->Put(write_opts, Key1(key), ts_str, "foo_" + std::to_string(key)));
}
// Write another L0 with keys 99 with newer ts.
ASSERT_OK(Flush());
uint64_t saved_read_ts1 = ts++;
key = 99;
for (int i = 0; i < 4; ++i, ++ts) {
std::string ts_str = Timestamp(ts);
Revise APIs related to user-defined timestamp (#8946) Summary: ajkr reminded me that we have a rule of not including per-kv related data in `WriteOptions`. Namely, `WriteOptions` should not include information about "what-to-write", but should just include information about "how-to-write". According to this rule, `WriteOptions::timestamp` (experimental) is clearly a violation. Therefore, this PR removes `WriteOptions::timestamp` for compliance. After the removal, we need to pass timestamp info via another set of APIs. This PR proposes a set of overloaded functions `Put(write_opts, key, value, ts)`, `Delete(write_opts, key, ts)`, and `SingleDelete(write_opts, key, ts)`. Planned to add `Write(write_opts, batch, ts)`, but its complexity made me reconsider doing it in another PR (maybe). For better checking and returning error early, we also add a new set of APIs to `WriteBatch` that take extra `timestamp` information when writing to `WriteBatch`es. These set of APIs in `WriteBatchWithIndex` are currently not supported, and are on our TODO list. Removed `WriteBatch::AssignTimestamps()` and renamed `WriteBatch::AssignTimestamp()` to `WriteBatch::UpdateTimestamps()` since this method require that all keys have space for timestamps allocated already and multiple timestamps can be updated. The constructor of `WriteBatch` now takes a fourth argument `default_cf_ts_sz` which is the timestamp size of the default column family. This will be used to allocate space when calling APIs that do not specify a column family handle. Also, updated `DB::Get()`, `DB::MultiGet()`, `DB::NewIterator()`, `DB::NewIterators()` methods, replacing some assertions about timestamp to returning Status code. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8946 Test Plan: make check ./db_bench -benchmarks=fillseq,fillrandom,readrandom,readseq,deleterandom -user_timestamp_size=8 ./db_stress --user_timestamp_size=8 -nooverwritepercent=0 -test_secondary=0 -secondary_catch_up_one_in=0 -continuous_verification_interval=0 Make sure there is no perf regression by running the following ``` ./db_bench_opt -db=/dev/shm/rocksdb -use_existing_db=0 -level0_stop_writes_trigger=256 -level0_slowdown_writes_trigger=256 -level0_file_num_compaction_trigger=256 -disable_wal=1 -duration=10 -benchmarks=fillrandom ``` Before this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.831 micros/op 546235 ops/sec; 60.4 MB/s ``` After this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.820 micros/op 549404 ops/sec; 60.8 MB/s ``` Reviewed By: ltamasi Differential Revision: D33721359 Pulled By: riversand963 fbshipit-source-id: c131561534272c120ffb80711d42748d21badf09
3 years ago
ASSERT_OK(
db_->Put(write_opts, Key1(key), ts_str, "bar_" + std::to_string(key)));
}
ASSERT_OK(Flush());
uint64_t saved_read_ts2 = ts++;
// Write another L0 with keys 99, 100, 101, ..., 150
for (; key <= 150; ++key, ++ts) {
std::string ts_str = Timestamp(ts);
Revise APIs related to user-defined timestamp (#8946) Summary: ajkr reminded me that we have a rule of not including per-kv related data in `WriteOptions`. Namely, `WriteOptions` should not include information about "what-to-write", but should just include information about "how-to-write". According to this rule, `WriteOptions::timestamp` (experimental) is clearly a violation. Therefore, this PR removes `WriteOptions::timestamp` for compliance. After the removal, we need to pass timestamp info via another set of APIs. This PR proposes a set of overloaded functions `Put(write_opts, key, value, ts)`, `Delete(write_opts, key, ts)`, and `SingleDelete(write_opts, key, ts)`. Planned to add `Write(write_opts, batch, ts)`, but its complexity made me reconsider doing it in another PR (maybe). For better checking and returning error early, we also add a new set of APIs to `WriteBatch` that take extra `timestamp` information when writing to `WriteBatch`es. These set of APIs in `WriteBatchWithIndex` are currently not supported, and are on our TODO list. Removed `WriteBatch::AssignTimestamps()` and renamed `WriteBatch::AssignTimestamp()` to `WriteBatch::UpdateTimestamps()` since this method require that all keys have space for timestamps allocated already and multiple timestamps can be updated. The constructor of `WriteBatch` now takes a fourth argument `default_cf_ts_sz` which is the timestamp size of the default column family. This will be used to allocate space when calling APIs that do not specify a column family handle. Also, updated `DB::Get()`, `DB::MultiGet()`, `DB::NewIterator()`, `DB::NewIterators()` methods, replacing some assertions about timestamp to returning Status code. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8946 Test Plan: make check ./db_bench -benchmarks=fillseq,fillrandom,readrandom,readseq,deleterandom -user_timestamp_size=8 ./db_stress --user_timestamp_size=8 -nooverwritepercent=0 -test_secondary=0 -secondary_catch_up_one_in=0 -continuous_verification_interval=0 Make sure there is no perf regression by running the following ``` ./db_bench_opt -db=/dev/shm/rocksdb -use_existing_db=0 -level0_stop_writes_trigger=256 -level0_slowdown_writes_trigger=256 -level0_file_num_compaction_trigger=256 -disable_wal=1 -duration=10 -benchmarks=fillrandom ``` Before this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.831 micros/op 546235 ops/sec; 60.4 MB/s ``` After this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.820 micros/op 549404 ops/sec; 60.8 MB/s ``` Reviewed By: ltamasi Differential Revision: D33721359 Pulled By: riversand963 fbshipit-source-id: c131561534272c120ffb80711d42748d21badf09
3 years ago
ASSERT_OK(
db_->Put(write_opts, Key1(key), ts_str, "foo1_" + std::to_string(key)));
}
ASSERT_OK(Flush());
// Wait for compaction to finish
ASSERT_OK(dbfull()->TEST_WaitForCompact());
uint64_t read_ts = ts;
ASSERT_EQ("foo_99", Get(Key1(99), saved_read_ts1));
ASSERT_EQ("bar_99", Get(Key1(99), saved_read_ts2));
ASSERT_EQ("foo1_99", Get(Key1(99), read_ts));
SyncPoint::GetInstance()->ClearAllCallBacks();
SyncPoint::GetInstance()->DisableProcessing();
}
Provide support for subcompactions with user-defined timestamps (#10344) Summary: The subcompaction logic currently picks file boundaries as subcompaction boundaries. This is not compatible with user-defined timestamps because of two issues. Issue1: ReadOptions.iterate_lower_bound and ReadOptions.iterate_upper_bound contains timestamps which results in assertion failure as BlockBasedTableIterator expects bounds to be without timestamps. As result, because of wrong comparison end key is returned as user_key resulting in assertion failure. Issue2: Since it might result in two keys that only differ by user timestamp getting processed by two different subcompactions (and thus two different CompactionIterator state machines), which in turn can cause data correction issues. This PR provide support to reenable subcompactions with user-defined timestamps. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10344 Test Plan: Added new unit test - Without fix for Issue1 unit test MultipleSubCompactions fails with error: ``` db_with_timestamp_compaction_test: ./db/compaction/clipping_iterator.h:247: void rocksdb::ClippingIterat│ or::AssertBounds(): Assertion `!valid_ || !end_ || cmp_->Compare(key(), *end_) < 0' failed. Received signal 6 (Aborted) │ #0 /usr/local/fbcode/platform009/lib/libc.so.6(gsignal+0x100) [0x7f8fbbbfe530] db_with_timestamp_compaction_test: ./db/compaction/clipping_iterator.h:247: void rocksdb::ClippingIterator::AssertBounds(): Assertion `!valid_ || !end_ || cmp_->Compare(key(), *end_) < 0' failed. Aborted (core dumped) ``` Ran stress test `make crash_test_with_ts -j32` Reviewed By: riversand963 Differential Revision: D38220841 Pulled By: akankshamahajan15 fbshipit-source-id: 5d5cae2bd37fcaeba1e77fce0a69070ad4158ccb
2 years ago
TEST_F(TimestampCompatibleCompactionTest, MultipleSubCompactions) {
Options options = CurrentOptions();
options.env = env_;
options.compaction_style = kCompactionStyleUniversal;
options.comparator = test::BytewiseComparatorWithU64TsWrapper();
options.level0_file_num_compaction_trigger = 3;
options.max_subcompactions = 3;
options.target_file_size_base = 1024;
options.statistics = CreateDBStatistics();
DestroyAndReopen(options);
uint64_t ts = 100;
uint64_t key = 0;
WriteOptions write_opts;
// Write keys 0, 1, ..., 499 with ts from 100 to 599.
{
for (; key <= 499; ++key, ++ts) {
std::string ts_str = Timestamp(ts);
ASSERT_OK(db_->Put(write_opts, Key1(key), ts_str,
"foo_" + std::to_string(key)));
}
}
// Write keys 500, ..., 999 with ts from 600 to 1099.
{
for (; key <= 999; ++key, ++ts) {
std::string ts_str = Timestamp(ts);
ASSERT_OK(db_->Put(write_opts, Key1(key), ts_str,
"foo_" + std::to_string(key)));
}
ASSERT_OK(Flush());
}
// Wait for compaction to finish
{
ASSERT_OK(dbfull()->RunManualCompaction(
static_cast_with_check<ColumnFamilyHandleImpl>(
db_->DefaultColumnFamily())
->cfd(),
0 /* input_level */, 1 /* output_level */, CompactRangeOptions(),
nullptr /* begin */, nullptr /* end */, true /* exclusive */,
true /* disallow_trivial_move */,
std::numeric_limits<uint64_t>::max() /* max_file_num_to_ignore */,
"" /*trim_ts*/));
}
// Check stats to make sure multiple subcompactions were scheduled for
// boundaries not to be nullptr.
{
HistogramData num_sub_compactions;
options.statistics->histogramData(NUM_SUBCOMPACTIONS_SCHEDULED,
&num_sub_compactions);
ASSERT_GT(num_sub_compactions.sum, 1);
}
for (key = 0; key <= 999; ++key) {
ASSERT_EQ("foo_" + std::to_string(key), Get(Key1(key), ts));
}
}
} // namespace ROCKSDB_NAMESPACE
int main(int argc, char** argv) {
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}