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/compaction/compaction_picker.cc

1228 lines
47 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_picker.h"
#include <cinttypes>
#include <limits>
#include <queue>
#include <string>
#include <utility>
#include <vector>
#include "db/column_family.h"
#include "file/filename.h"
#include "logging/log_buffer.h"
#include "logging/logging.h"
#include "monitoring/statistics.h"
#include "test_util/sync_point.h"
#include "util/random.h"
#include "util/string_util.h"
namespace ROCKSDB_NAMESPACE {
bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
size_t min_files_to_compact,
uint64_t max_compact_bytes_per_del_file,
uint64_t max_compaction_bytes,
Sort L0 files by newly introduced epoch_num (#10922) Summary: **Context:** Sorting L0 files by `largest_seqno` has at least two inconvenience: - File ingestion and compaction involving ingested files can create files of overlapping seqno range with the existing files. `force_consistency_check=true` will catch such overlap seqno range even those harmless overlap. - For example, consider the following sequence of events ("key@n" indicates key at seqno "n") - insert k1@1 to memtable m1 - ingest file s1 with k2@2, ingest file s2 with k3@3 - insert k4@4 to m1 - compact files s1, s2 and result in new file s3 of seqno range [2, 3] - flush m1 and result in new file s4 of seqno range [1, 4]. And `force_consistency_check=true` will think s4 and s3 has file reordering corruption that might cause retuning an old value of k1 - However such caught corruption is a false positive since s1, s2 will not have overlapped keys with k1 or whatever inserted into m1 before ingest file s1 by the requirement of file ingestion (otherwise the m1 will be flushed first before any of the file ingestion completes). Therefore there in fact isn't any file reordering corruption. - Single delete can decrease a file's largest seqno and ordering by `largest_seqno` can introduce a wrong ordering hence file reordering corruption - For example, consider the following sequence of events ("key@n" indicates key at seqno "n", Credit to ajkr for this example) - an existing SST s1 contains only k1@1 - insert k1@2 to memtable m1 - ingest file s2 with k3@3, ingest file s3 with k4@4 - insert single delete k5@5 in m1 - flush m1 and result in new file s4 of seqno range [2, 5] - compact s1, s2, s3 and result in new file s5 of seqno range [1, 4] - compact s4 and result in new file s6 of seqno range [2] due to single delete - By the last step, we have file ordering by largest seqno (">" means "newer") : s5 > s6 while s6 contains a newer version of the k1's value (i.e, k1@2) than s5, which is a real reordering corruption. While this can be caught by `force_consistency_check=true`, there isn't a good way to prevent this from happening if ordering by `largest_seqno` Therefore, we are redesigning the sorting criteria of L0 files and avoid above inconvenience. Credit to ajkr , we now introduce `epoch_num` which describes the order of a file being flushed or ingested/imported (compaction output file will has the minimum `epoch_num` among input files'). This will avoid the above inconvenience in the following ways: - In the first case above, there will no longer be overlap seqno range check in `force_consistency_check=true` but `epoch_number` ordering check. This will result in file ordering s1 < s2 < s4 (pre-compaction) and s3 < s4 (post-compaction) which won't trigger false positive corruption. See test class `DBCompactionTestL0FilesMisorderCorruption*` for more. - In the second case above, this will result in file ordering s1 < s2 < s3 < s4 (pre-compacting s1, s2, s3), s5 < s4 (post-compacting s1, s2, s3), s5 < s6 (post-compacting s4), which are correct file ordering without causing any corruption. **Summary:** - Introduce `epoch_number` stored per `ColumnFamilyData` and sort CF's L0 files by their assigned `epoch_number` instead of `largest_seqno`. - `epoch_number` is increased and assigned upon `VersionEdit::AddFile()` for flush (or similarly for WriteLevel0TableForRecovery) and file ingestion (except for allow_behind_true, which will always get assigned as the `kReservedEpochNumberForFileIngestedBehind`) - Compaction output file is assigned with the minimum `epoch_number` among input files' - Refit level: reuse refitted file's epoch_number - Other paths needing `epoch_number` treatment: - Import column families: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo` - Repair: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo`. - Assigning new epoch_number to a file and adding this file to LSM tree should be atomic. This is guaranteed by us assigning epoch_number right upon `VersionEdit::AddFile()` where this version edit will be apply to LSM tree shape right after by holding the db mutex (e.g, flush, file ingestion, import column family) or by there is only 1 ongoing edit per CF (e.g, WriteLevel0TableForRecovery, Repair). - Assigning the minimum input epoch number to compaction output file won't misorder L0 files (even through later `Refit(target_level=0)`). It's due to for every key "k" in the input range, a legit compaction will cover a continuous epoch number range of that key. As long as we assign the key "k" the minimum input epoch number, it won't become newer or older than the versions of this key that aren't included in this compaction hence no misorder. - Persist `epoch_number` of each file in manifest and recover `epoch_number` on db recovery - Backward compatibility with old db without `epoch_number` support is guaranteed by assigning `epoch_number` to recovered files by `NewestFirstBySeqno` order. See `VersionStorageInfo::RecoverEpochNumbers()` for more - Forward compatibility with manifest is guaranteed by flexibility of `NewFileCustomTag` - Replace `force_consistent_check` on L0 with `epoch_number` and remove false positive check like case 1 with `largest_seqno` above - Due to backward compatibility issue, we might encounter files with missing epoch number at the beginning of db recovery. We will still use old L0 sorting mechanism (`NewestFirstBySeqno`) to check/sort them till we infer their epoch number. See usages of `EpochNumberRequirement`. - Remove fix https://github.com/facebook/rocksdb/pull/5958#issue-511150930 and their outdated tests to file reordering corruption because such fix can be replaced by this PR. - Misc: - update existing tests with `epoch_number` so make check will pass - update https://github.com/facebook/rocksdb/pull/5958#issue-511150930 tests to verify corruption is fixed using `epoch_number` and cover universal/fifo compaction/CompactRange/CompactFile cases - assert db_mutex is held for a few places before calling ColumnFamilyData::NewEpochNumber() Pull Request resolved: https://github.com/facebook/rocksdb/pull/10922 Test Plan: - `make check` - New unit tests under `db/db_compaction_test.cc`, `db/db_test2.cc`, `db/version_builder_test.cc`, `db/repair_test.cc` - Updated tests (i.e, `DBCompactionTestL0FilesMisorderCorruption*`) under https://github.com/facebook/rocksdb/pull/5958#issue-511150930 - [Ongoing] Compatibility test: manually run https://github.com/ajkr/rocksdb/commit/36a5686ec012f35a4371e409aa85c404ca1c210d (with file ingestion off for running the `.orig` binary to prevent this bug affecting upgrade/downgrade formality checking) for 1 hour on `simple black/white box`, `cf_consistency/txn/enable_ts with whitebox + test_best_efforts_recovery with blackbox` - [Ongoing] normal db stress test - [Ongoing] db stress test with aggressive value https://github.com/facebook/rocksdb/pull/10761 Reviewed By: ajkr Differential Revision: D41063187 Pulled By: hx235 fbshipit-source-id: 826cb23455de7beaabe2d16c57682a82733a32a9
2 years ago
CompactionInputFiles* comp_inputs) {
TEST_SYNC_POINT("FindIntraL0Compaction");
Sort L0 files by newly introduced epoch_num (#10922) Summary: **Context:** Sorting L0 files by `largest_seqno` has at least two inconvenience: - File ingestion and compaction involving ingested files can create files of overlapping seqno range with the existing files. `force_consistency_check=true` will catch such overlap seqno range even those harmless overlap. - For example, consider the following sequence of events ("key@n" indicates key at seqno "n") - insert k1@1 to memtable m1 - ingest file s1 with k2@2, ingest file s2 with k3@3 - insert k4@4 to m1 - compact files s1, s2 and result in new file s3 of seqno range [2, 3] - flush m1 and result in new file s4 of seqno range [1, 4]. And `force_consistency_check=true` will think s4 and s3 has file reordering corruption that might cause retuning an old value of k1 - However such caught corruption is a false positive since s1, s2 will not have overlapped keys with k1 or whatever inserted into m1 before ingest file s1 by the requirement of file ingestion (otherwise the m1 will be flushed first before any of the file ingestion completes). Therefore there in fact isn't any file reordering corruption. - Single delete can decrease a file's largest seqno and ordering by `largest_seqno` can introduce a wrong ordering hence file reordering corruption - For example, consider the following sequence of events ("key@n" indicates key at seqno "n", Credit to ajkr for this example) - an existing SST s1 contains only k1@1 - insert k1@2 to memtable m1 - ingest file s2 with k3@3, ingest file s3 with k4@4 - insert single delete k5@5 in m1 - flush m1 and result in new file s4 of seqno range [2, 5] - compact s1, s2, s3 and result in new file s5 of seqno range [1, 4] - compact s4 and result in new file s6 of seqno range [2] due to single delete - By the last step, we have file ordering by largest seqno (">" means "newer") : s5 > s6 while s6 contains a newer version of the k1's value (i.e, k1@2) than s5, which is a real reordering corruption. While this can be caught by `force_consistency_check=true`, there isn't a good way to prevent this from happening if ordering by `largest_seqno` Therefore, we are redesigning the sorting criteria of L0 files and avoid above inconvenience. Credit to ajkr , we now introduce `epoch_num` which describes the order of a file being flushed or ingested/imported (compaction output file will has the minimum `epoch_num` among input files'). This will avoid the above inconvenience in the following ways: - In the first case above, there will no longer be overlap seqno range check in `force_consistency_check=true` but `epoch_number` ordering check. This will result in file ordering s1 < s2 < s4 (pre-compaction) and s3 < s4 (post-compaction) which won't trigger false positive corruption. See test class `DBCompactionTestL0FilesMisorderCorruption*` for more. - In the second case above, this will result in file ordering s1 < s2 < s3 < s4 (pre-compacting s1, s2, s3), s5 < s4 (post-compacting s1, s2, s3), s5 < s6 (post-compacting s4), which are correct file ordering without causing any corruption. **Summary:** - Introduce `epoch_number` stored per `ColumnFamilyData` and sort CF's L0 files by their assigned `epoch_number` instead of `largest_seqno`. - `epoch_number` is increased and assigned upon `VersionEdit::AddFile()` for flush (or similarly for WriteLevel0TableForRecovery) and file ingestion (except for allow_behind_true, which will always get assigned as the `kReservedEpochNumberForFileIngestedBehind`) - Compaction output file is assigned with the minimum `epoch_number` among input files' - Refit level: reuse refitted file's epoch_number - Other paths needing `epoch_number` treatment: - Import column families: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo` - Repair: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo`. - Assigning new epoch_number to a file and adding this file to LSM tree should be atomic. This is guaranteed by us assigning epoch_number right upon `VersionEdit::AddFile()` where this version edit will be apply to LSM tree shape right after by holding the db mutex (e.g, flush, file ingestion, import column family) or by there is only 1 ongoing edit per CF (e.g, WriteLevel0TableForRecovery, Repair). - Assigning the minimum input epoch number to compaction output file won't misorder L0 files (even through later `Refit(target_level=0)`). It's due to for every key "k" in the input range, a legit compaction will cover a continuous epoch number range of that key. As long as we assign the key "k" the minimum input epoch number, it won't become newer or older than the versions of this key that aren't included in this compaction hence no misorder. - Persist `epoch_number` of each file in manifest and recover `epoch_number` on db recovery - Backward compatibility with old db without `epoch_number` support is guaranteed by assigning `epoch_number` to recovered files by `NewestFirstBySeqno` order. See `VersionStorageInfo::RecoverEpochNumbers()` for more - Forward compatibility with manifest is guaranteed by flexibility of `NewFileCustomTag` - Replace `force_consistent_check` on L0 with `epoch_number` and remove false positive check like case 1 with `largest_seqno` above - Due to backward compatibility issue, we might encounter files with missing epoch number at the beginning of db recovery. We will still use old L0 sorting mechanism (`NewestFirstBySeqno`) to check/sort them till we infer their epoch number. See usages of `EpochNumberRequirement`. - Remove fix https://github.com/facebook/rocksdb/pull/5958#issue-511150930 and their outdated tests to file reordering corruption because such fix can be replaced by this PR. - Misc: - update existing tests with `epoch_number` so make check will pass - update https://github.com/facebook/rocksdb/pull/5958#issue-511150930 tests to verify corruption is fixed using `epoch_number` and cover universal/fifo compaction/CompactRange/CompactFile cases - assert db_mutex is held for a few places before calling ColumnFamilyData::NewEpochNumber() Pull Request resolved: https://github.com/facebook/rocksdb/pull/10922 Test Plan: - `make check` - New unit tests under `db/db_compaction_test.cc`, `db/db_test2.cc`, `db/version_builder_test.cc`, `db/repair_test.cc` - Updated tests (i.e, `DBCompactionTestL0FilesMisorderCorruption*`) under https://github.com/facebook/rocksdb/pull/5958#issue-511150930 - [Ongoing] Compatibility test: manually run https://github.com/ajkr/rocksdb/commit/36a5686ec012f35a4371e409aa85c404ca1c210d (with file ingestion off for running the `.orig` binary to prevent this bug affecting upgrade/downgrade formality checking) for 1 hour on `simple black/white box`, `cf_consistency/txn/enable_ts with whitebox + test_best_efforts_recovery with blackbox` - [Ongoing] normal db stress test - [Ongoing] db stress test with aggressive value https://github.com/facebook/rocksdb/pull/10761 Reviewed By: ajkr Differential Revision: D41063187 Pulled By: hx235 fbshipit-source-id: 826cb23455de7beaabe2d16c57682a82733a32a9
2 years ago
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
size_t start = 0;
Sort L0 files by newly introduced epoch_num (#10922) Summary: **Context:** Sorting L0 files by `largest_seqno` has at least two inconvenience: - File ingestion and compaction involving ingested files can create files of overlapping seqno range with the existing files. `force_consistency_check=true` will catch such overlap seqno range even those harmless overlap. - For example, consider the following sequence of events ("key@n" indicates key at seqno "n") - insert k1@1 to memtable m1 - ingest file s1 with k2@2, ingest file s2 with k3@3 - insert k4@4 to m1 - compact files s1, s2 and result in new file s3 of seqno range [2, 3] - flush m1 and result in new file s4 of seqno range [1, 4]. And `force_consistency_check=true` will think s4 and s3 has file reordering corruption that might cause retuning an old value of k1 - However such caught corruption is a false positive since s1, s2 will not have overlapped keys with k1 or whatever inserted into m1 before ingest file s1 by the requirement of file ingestion (otherwise the m1 will be flushed first before any of the file ingestion completes). Therefore there in fact isn't any file reordering corruption. - Single delete can decrease a file's largest seqno and ordering by `largest_seqno` can introduce a wrong ordering hence file reordering corruption - For example, consider the following sequence of events ("key@n" indicates key at seqno "n", Credit to ajkr for this example) - an existing SST s1 contains only k1@1 - insert k1@2 to memtable m1 - ingest file s2 with k3@3, ingest file s3 with k4@4 - insert single delete k5@5 in m1 - flush m1 and result in new file s4 of seqno range [2, 5] - compact s1, s2, s3 and result in new file s5 of seqno range [1, 4] - compact s4 and result in new file s6 of seqno range [2] due to single delete - By the last step, we have file ordering by largest seqno (">" means "newer") : s5 > s6 while s6 contains a newer version of the k1's value (i.e, k1@2) than s5, which is a real reordering corruption. While this can be caught by `force_consistency_check=true`, there isn't a good way to prevent this from happening if ordering by `largest_seqno` Therefore, we are redesigning the sorting criteria of L0 files and avoid above inconvenience. Credit to ajkr , we now introduce `epoch_num` which describes the order of a file being flushed or ingested/imported (compaction output file will has the minimum `epoch_num` among input files'). This will avoid the above inconvenience in the following ways: - In the first case above, there will no longer be overlap seqno range check in `force_consistency_check=true` but `epoch_number` ordering check. This will result in file ordering s1 < s2 < s4 (pre-compaction) and s3 < s4 (post-compaction) which won't trigger false positive corruption. See test class `DBCompactionTestL0FilesMisorderCorruption*` for more. - In the second case above, this will result in file ordering s1 < s2 < s3 < s4 (pre-compacting s1, s2, s3), s5 < s4 (post-compacting s1, s2, s3), s5 < s6 (post-compacting s4), which are correct file ordering without causing any corruption. **Summary:** - Introduce `epoch_number` stored per `ColumnFamilyData` and sort CF's L0 files by their assigned `epoch_number` instead of `largest_seqno`. - `epoch_number` is increased and assigned upon `VersionEdit::AddFile()` for flush (or similarly for WriteLevel0TableForRecovery) and file ingestion (except for allow_behind_true, which will always get assigned as the `kReservedEpochNumberForFileIngestedBehind`) - Compaction output file is assigned with the minimum `epoch_number` among input files' - Refit level: reuse refitted file's epoch_number - Other paths needing `epoch_number` treatment: - Import column families: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo` - Repair: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo`. - Assigning new epoch_number to a file and adding this file to LSM tree should be atomic. This is guaranteed by us assigning epoch_number right upon `VersionEdit::AddFile()` where this version edit will be apply to LSM tree shape right after by holding the db mutex (e.g, flush, file ingestion, import column family) or by there is only 1 ongoing edit per CF (e.g, WriteLevel0TableForRecovery, Repair). - Assigning the minimum input epoch number to compaction output file won't misorder L0 files (even through later `Refit(target_level=0)`). It's due to for every key "k" in the input range, a legit compaction will cover a continuous epoch number range of that key. As long as we assign the key "k" the minimum input epoch number, it won't become newer or older than the versions of this key that aren't included in this compaction hence no misorder. - Persist `epoch_number` of each file in manifest and recover `epoch_number` on db recovery - Backward compatibility with old db without `epoch_number` support is guaranteed by assigning `epoch_number` to recovered files by `NewestFirstBySeqno` order. See `VersionStorageInfo::RecoverEpochNumbers()` for more - Forward compatibility with manifest is guaranteed by flexibility of `NewFileCustomTag` - Replace `force_consistent_check` on L0 with `epoch_number` and remove false positive check like case 1 with `largest_seqno` above - Due to backward compatibility issue, we might encounter files with missing epoch number at the beginning of db recovery. We will still use old L0 sorting mechanism (`NewestFirstBySeqno`) to check/sort them till we infer their epoch number. See usages of `EpochNumberRequirement`. - Remove fix https://github.com/facebook/rocksdb/pull/5958#issue-511150930 and their outdated tests to file reordering corruption because such fix can be replaced by this PR. - Misc: - update existing tests with `epoch_number` so make check will pass - update https://github.com/facebook/rocksdb/pull/5958#issue-511150930 tests to verify corruption is fixed using `epoch_number` and cover universal/fifo compaction/CompactRange/CompactFile cases - assert db_mutex is held for a few places before calling ColumnFamilyData::NewEpochNumber() Pull Request resolved: https://github.com/facebook/rocksdb/pull/10922 Test Plan: - `make check` - New unit tests under `db/db_compaction_test.cc`, `db/db_test2.cc`, `db/version_builder_test.cc`, `db/repair_test.cc` - Updated tests (i.e, `DBCompactionTestL0FilesMisorderCorruption*`) under https://github.com/facebook/rocksdb/pull/5958#issue-511150930 - [Ongoing] Compatibility test: manually run https://github.com/ajkr/rocksdb/commit/36a5686ec012f35a4371e409aa85c404ca1c210d (with file ingestion off for running the `.orig` binary to prevent this bug affecting upgrade/downgrade formality checking) for 1 hour on `simple black/white box`, `cf_consistency/txn/enable_ts with whitebox + test_best_efforts_recovery with blackbox` - [Ongoing] normal db stress test - [Ongoing] db stress test with aggressive value https://github.com/facebook/rocksdb/pull/10761 Reviewed By: ajkr Differential Revision: D41063187 Pulled By: hx235 fbshipit-source-id: 826cb23455de7beaabe2d16c57682a82733a32a9
2 years ago
if (level_files.size() == 0 || level_files[start]->being_compacted) {
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
return false;
}
Sort L0 files by newly introduced epoch_num (#10922) Summary: **Context:** Sorting L0 files by `largest_seqno` has at least two inconvenience: - File ingestion and compaction involving ingested files can create files of overlapping seqno range with the existing files. `force_consistency_check=true` will catch such overlap seqno range even those harmless overlap. - For example, consider the following sequence of events ("key@n" indicates key at seqno "n") - insert k1@1 to memtable m1 - ingest file s1 with k2@2, ingest file s2 with k3@3 - insert k4@4 to m1 - compact files s1, s2 and result in new file s3 of seqno range [2, 3] - flush m1 and result in new file s4 of seqno range [1, 4]. And `force_consistency_check=true` will think s4 and s3 has file reordering corruption that might cause retuning an old value of k1 - However such caught corruption is a false positive since s1, s2 will not have overlapped keys with k1 or whatever inserted into m1 before ingest file s1 by the requirement of file ingestion (otherwise the m1 will be flushed first before any of the file ingestion completes). Therefore there in fact isn't any file reordering corruption. - Single delete can decrease a file's largest seqno and ordering by `largest_seqno` can introduce a wrong ordering hence file reordering corruption - For example, consider the following sequence of events ("key@n" indicates key at seqno "n", Credit to ajkr for this example) - an existing SST s1 contains only k1@1 - insert k1@2 to memtable m1 - ingest file s2 with k3@3, ingest file s3 with k4@4 - insert single delete k5@5 in m1 - flush m1 and result in new file s4 of seqno range [2, 5] - compact s1, s2, s3 and result in new file s5 of seqno range [1, 4] - compact s4 and result in new file s6 of seqno range [2] due to single delete - By the last step, we have file ordering by largest seqno (">" means "newer") : s5 > s6 while s6 contains a newer version of the k1's value (i.e, k1@2) than s5, which is a real reordering corruption. While this can be caught by `force_consistency_check=true`, there isn't a good way to prevent this from happening if ordering by `largest_seqno` Therefore, we are redesigning the sorting criteria of L0 files and avoid above inconvenience. Credit to ajkr , we now introduce `epoch_num` which describes the order of a file being flushed or ingested/imported (compaction output file will has the minimum `epoch_num` among input files'). This will avoid the above inconvenience in the following ways: - In the first case above, there will no longer be overlap seqno range check in `force_consistency_check=true` but `epoch_number` ordering check. This will result in file ordering s1 < s2 < s4 (pre-compaction) and s3 < s4 (post-compaction) which won't trigger false positive corruption. See test class `DBCompactionTestL0FilesMisorderCorruption*` for more. - In the second case above, this will result in file ordering s1 < s2 < s3 < s4 (pre-compacting s1, s2, s3), s5 < s4 (post-compacting s1, s2, s3), s5 < s6 (post-compacting s4), which are correct file ordering without causing any corruption. **Summary:** - Introduce `epoch_number` stored per `ColumnFamilyData` and sort CF's L0 files by their assigned `epoch_number` instead of `largest_seqno`. - `epoch_number` is increased and assigned upon `VersionEdit::AddFile()` for flush (or similarly for WriteLevel0TableForRecovery) and file ingestion (except for allow_behind_true, which will always get assigned as the `kReservedEpochNumberForFileIngestedBehind`) - Compaction output file is assigned with the minimum `epoch_number` among input files' - Refit level: reuse refitted file's epoch_number - Other paths needing `epoch_number` treatment: - Import column families: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo` - Repair: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo`. - Assigning new epoch_number to a file and adding this file to LSM tree should be atomic. This is guaranteed by us assigning epoch_number right upon `VersionEdit::AddFile()` where this version edit will be apply to LSM tree shape right after by holding the db mutex (e.g, flush, file ingestion, import column family) or by there is only 1 ongoing edit per CF (e.g, WriteLevel0TableForRecovery, Repair). - Assigning the minimum input epoch number to compaction output file won't misorder L0 files (even through later `Refit(target_level=0)`). It's due to for every key "k" in the input range, a legit compaction will cover a continuous epoch number range of that key. As long as we assign the key "k" the minimum input epoch number, it won't become newer or older than the versions of this key that aren't included in this compaction hence no misorder. - Persist `epoch_number` of each file in manifest and recover `epoch_number` on db recovery - Backward compatibility with old db without `epoch_number` support is guaranteed by assigning `epoch_number` to recovered files by `NewestFirstBySeqno` order. See `VersionStorageInfo::RecoverEpochNumbers()` for more - Forward compatibility with manifest is guaranteed by flexibility of `NewFileCustomTag` - Replace `force_consistent_check` on L0 with `epoch_number` and remove false positive check like case 1 with `largest_seqno` above - Due to backward compatibility issue, we might encounter files with missing epoch number at the beginning of db recovery. We will still use old L0 sorting mechanism (`NewestFirstBySeqno`) to check/sort them till we infer their epoch number. See usages of `EpochNumberRequirement`. - Remove fix https://github.com/facebook/rocksdb/pull/5958#issue-511150930 and their outdated tests to file reordering corruption because such fix can be replaced by this PR. - Misc: - update existing tests with `epoch_number` so make check will pass - update https://github.com/facebook/rocksdb/pull/5958#issue-511150930 tests to verify corruption is fixed using `epoch_number` and cover universal/fifo compaction/CompactRange/CompactFile cases - assert db_mutex is held for a few places before calling ColumnFamilyData::NewEpochNumber() Pull Request resolved: https://github.com/facebook/rocksdb/pull/10922 Test Plan: - `make check` - New unit tests under `db/db_compaction_test.cc`, `db/db_test2.cc`, `db/version_builder_test.cc`, `db/repair_test.cc` - Updated tests (i.e, `DBCompactionTestL0FilesMisorderCorruption*`) under https://github.com/facebook/rocksdb/pull/5958#issue-511150930 - [Ongoing] Compatibility test: manually run https://github.com/ajkr/rocksdb/commit/36a5686ec012f35a4371e409aa85c404ca1c210d (with file ingestion off for running the `.orig` binary to prevent this bug affecting upgrade/downgrade formality checking) for 1 hour on `simple black/white box`, `cf_consistency/txn/enable_ts with whitebox + test_best_efforts_recovery with blackbox` - [Ongoing] normal db stress test - [Ongoing] db stress test with aggressive value https://github.com/facebook/rocksdb/pull/10761 Reviewed By: ajkr Differential Revision: D41063187 Pulled By: hx235 fbshipit-source-id: 826cb23455de7beaabe2d16c57682a82733a32a9
2 years ago
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
size_t compact_bytes = static_cast<size_t>(level_files[start]->fd.file_size);
size_t compact_bytes_per_del_file = std::numeric_limits<size_t>::max();
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
// Compaction range will be [start, limit).
size_t limit;
// Pull in files until the amount of compaction work per deleted file begins
// increasing or maximum total compaction size is reached.
size_t new_compact_bytes_per_del_file = 0;
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
for (limit = start + 1; limit < level_files.size(); ++limit) {
compact_bytes += static_cast<size_t>(level_files[limit]->fd.file_size);
new_compact_bytes_per_del_file = compact_bytes / (limit - start);
if (level_files[limit]->being_compacted ||
new_compact_bytes_per_del_file > compact_bytes_per_del_file ||
compact_bytes > max_compaction_bytes) {
break;
}
compact_bytes_per_del_file = new_compact_bytes_per_del_file;
}
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
if ((limit - start) >= min_files_to_compact &&
compact_bytes_per_del_file < max_compact_bytes_per_del_file) {
assert(comp_inputs != nullptr);
comp_inputs->level = 0;
Fix corruption with intra-L0 on ingested files (#5958) Summary: ## Problem Description Our process was abort when it call `CheckConsistency`. And the information in `stderr` show that "`L0 files seqno 3001491972 3004797440 vs. 3002875611 3004524421` ". Here are the causes of the accident I investigated. * RocksDB will call `CheckConsistency` whenever `MANIFEST` file is update. It will check sequence number interval of every file, except files which were ingested. * When one file is ingested into RocksDB, it will be assigned the value of global sequence number, and the minimum and maximum seqno of this file are equal, which are both equal to global sequence number. * `CheckConsistency` determines whether the file is ingested by whether the smallest and largest seqno of an sstable file are equal. * If IntraL0Compaction picks one sst which was ingested just now and compacted it into another sst, the `smallest_seqno` of this new file will be smaller than his `largest_seqno`. * If more than one ingested file was ingested before memtable schedule flush, and they all compact into one new sstable file by `IntraL0Compaction`. The sequence interval of this new file will be included in the interval of the memtable. So `CheckConsistency` will return a `Corruption`. * If a sstable was ingested after the memtable was schedule to flush, which would assign a larger seqno to it than memtable. Then the file was compacted with other files (these files were all flushed before the memtable) in L0 into one file. This compaction start before the flush job of memtable start, but completed after the flush job finish. So this new file produced by the compaction (we call it s1) would have a larger interval of sequence number than the file produced by flush (we call it s2). **But there was still some data in s1 written into RocksDB before the s2, so it's possible that some data in s2 was cover by old data in s1.** Of course, it would also make a `Corruption` because of overlap of seqno. There is the relationship of the files: > s1.smallest_seqno < s2.smallest_seqno < s2.largest_seqno < s1.largest_seqno So I skip pick sst file which was ingested in function `FindIntraL0Compaction ` ## Reason Here is my bug report: https://github.com/facebook/rocksdb/issues/5913 There are two situations that can cause the check to fail. ### First situation: - First we ingest five external sst into Rocksdb, and they happened to be ingested in L0. and there had been some data in memtable, which make the smallest sequence number of memtable is less than which of sst that we ingest. - If there had been one compaction job which compacted sst from L0 to L1, `LevelCompactionPicker` would trigger a `IntraL0Compaction` which would compact this five sst from L0 to L0. We call this sst A, which was merged from five ingested sst. - Then some data was put into memtable, and memtable was flushed to L0. We called this sst B. - RocksDB check consistency , and find the `smallest_seqno` of B is less than that of A and crash. Because A was merged from five sst, the smallest sequence number of it was less than the biggest sequece number of itself, so RocksDB could not tell if A was produce by ingested. ### Secondary situaion - First we have flushed many sst in L0, we call them [s1, s2, s3]. - There is an immutable memtable request to be flushed, but because flush thread is busy, so it has not been picked. we call it m1. And at the moment, one sst is ingested into L0. We call it s4. Because s4 is ingested after m1 became immutable memtable, so it has a larger log sequence number than m1. - m1 is flushed in L0. because it is small, this flush job finish quickly. we call it s5. - [s1, s2, s3, s4] are compacted into one sst to L0, by IntraL0Compaction. We call it s6. - compacted 4@0 files to L0 - When s6 is added into manifest, the corruption happened. because the largest sequence number of s6 is equal to s4, and they are both larger than that of s5. But because s1 is older than m1, so the smallest sequence number of s6 is smaller than that of s5. - s6.smallest_seqno < s5.smallest_seqno < s5.largest_seqno < s6.largest_seqno Pull Request resolved: https://github.com/facebook/rocksdb/pull/5958 Differential Revision: D18601316 fbshipit-source-id: 5fe54b3c9af52a2e1400728f565e895cde1c7267
5 years ago
for (size_t i = start; i < limit; ++i) {
comp_inputs->files.push_back(level_files[i]);
}
return true;
}
return false;
}
// Determine compression type, based on user options, level of the output
// file and whether compression is disabled.
// If enable_compression is false, then compression is always disabled no
// matter what the values of the other two parameters are.
// Otherwise, the compression type is determined based on options and level.
CompressionType GetCompressionType(const VersionStorageInfo* vstorage,
const MutableCFOptions& mutable_cf_options,
int level, int base_level,
const bool enable_compression) {
if (!enable_compression) {
// disable compression
return kNoCompression;
}
// If bottommost_compression is set and we are compacting to the
// bottommost level then we should use it.
if (mutable_cf_options.bottommost_compression != kDisableCompressionOption &&
level >= (vstorage->num_non_empty_levels() - 1)) {
return mutable_cf_options.bottommost_compression;
}
// If the user has specified a different compression level for each level,
// then pick the compression for that level.
if (!mutable_cf_options.compression_per_level.empty()) {
assert(level == 0 || level >= base_level);
int idx = (level == 0) ? 0 : level - base_level + 1;
const int n =
static_cast<int>(mutable_cf_options.compression_per_level.size()) - 1;
// It is possible for level_ to be -1; in that case, we use level
// 0's compression. This occurs mostly in backwards compatibility
// situations when the builder doesn't know what level the file
// belongs to. Likewise, if level is beyond the end of the
// specified compression levels, use the last value.
return mutable_cf_options
.compression_per_level[std::max(0, std::min(idx, n))];
} else {
return mutable_cf_options.compression;
}
}
CompressionOptions GetCompressionOptions(const MutableCFOptions& cf_options,
const VersionStorageInfo* vstorage,
int level,
const bool enable_compression) {
if (!enable_compression) {
return cf_options.compression_opts;
}
// If bottommost_compression_opts is enabled and we are compacting to the
// bottommost level then we should use the specified compression options.
if (level >= (vstorage->num_non_empty_levels() - 1) &&
cf_options.bottommost_compression_opts.enabled) {
return cf_options.bottommost_compression_opts;
}
return cf_options.compression_opts;
}
CompactionPicker::CompactionPicker(const ImmutableOptions& ioptions,
const InternalKeyComparator* icmp)
: ioptions_(ioptions), icmp_(icmp) {}
CompactionPicker::~CompactionPicker() {}
// Delete this compaction from the list of running compactions.
void CompactionPicker::ReleaseCompactionFiles(Compaction* c, Status status) {
UnregisterCompaction(c);
if (!status.ok()) {
c->ResetNextCompactionIndex();
}
}
void CompactionPicker::GetRange(const CompactionInputFiles& inputs,
InternalKey* smallest,
InternalKey* largest) const {
const int level = inputs.level;
assert(!inputs.empty());
smallest->Clear();
largest->Clear();
if (level == 0) {
for (size_t i = 0; i < inputs.size(); i++) {
FileMetaData* f = inputs[i];
if (i == 0) {
*smallest = f->smallest;
*largest = f->largest;
} else {
if (icmp_->Compare(f->smallest, *smallest) < 0) {
*smallest = f->smallest;
}
if (icmp_->Compare(f->largest, *largest) > 0) {
*largest = f->largest;
}
}
}
} else {
*smallest = inputs[0]->smallest;
*largest = inputs[inputs.size() - 1]->largest;
}
}
void CompactionPicker::GetRange(const CompactionInputFiles& inputs1,
const CompactionInputFiles& inputs2,
InternalKey* smallest,
InternalKey* largest) const {
assert(!inputs1.empty() || !inputs2.empty());
if (inputs1.empty()) {
GetRange(inputs2, smallest, largest);
} else if (inputs2.empty()) {
GetRange(inputs1, smallest, largest);
} else {
InternalKey smallest1, smallest2, largest1, largest2;
GetRange(inputs1, &smallest1, &largest1);
GetRange(inputs2, &smallest2, &largest2);
*smallest =
icmp_->Compare(smallest1, smallest2) < 0 ? smallest1 : smallest2;
*largest = icmp_->Compare(largest1, largest2) < 0 ? largest2 : largest1;
}
}
void CompactionPicker::GetRange(const std::vector<CompactionInputFiles>& inputs,
InternalKey* smallest, InternalKey* largest,
int exclude_level) const {
InternalKey current_smallest;
InternalKey current_largest;
bool initialized = false;
for (const auto& in : inputs) {
if (in.empty() || in.level == exclude_level) {
continue;
}
GetRange(in, &current_smallest, &current_largest);
if (!initialized) {
*smallest = current_smallest;
*largest = current_largest;
initialized = true;
} else {
if (icmp_->Compare(current_smallest, *smallest) < 0) {
*smallest = current_smallest;
}
if (icmp_->Compare(current_largest, *largest) > 0) {
*largest = current_largest;
}
}
}
assert(initialized);
}
bool CompactionPicker::ExpandInputsToCleanCut(const std::string& /*cf_name*/,
VersionStorageInfo* vstorage,
CompactionInputFiles* inputs,
InternalKey** next_smallest) {
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
// This isn't good compaction
assert(!inputs->empty());
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
const int level = inputs->level;
// GetOverlappingInputs will always do the right thing for level-0.
// So we don't need to do any expansion if level == 0.
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
if (level == 0) {
return true;
}
InternalKey smallest, largest;
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
// Keep expanding inputs until we are sure that there is a "clean cut"
// boundary between the files in input and the surrounding files.
// This will ensure that no parts of a key are lost during compaction.
int hint_index = -1;
size_t old_size;
do {
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
old_size = inputs->size();
GetRange(*inputs, &smallest, &largest);
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
inputs->clear();
vstorage->GetOverlappingInputs(level, &smallest, &largest, &inputs->files,
hint_index, &hint_index, true,
next_smallest);
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
} while (inputs->size() > old_size);
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
// we started off with inputs non-empty and the previous loop only grew
// inputs. thus, inputs should be non-empty here
assert(!inputs->empty());
// If, after the expansion, there are files that are already under
// compaction, then we must drop/cancel this compaction.
if (AreFilesInCompaction(inputs->files)) {
return false;
}
return true;
}
bool CompactionPicker::RangeOverlapWithCompaction(
const Slice& smallest_user_key, const Slice& largest_user_key,
int level) const {
const Comparator* ucmp = icmp_->user_comparator();
for (Compaction* c : compactions_in_progress_) {
if (c->output_level() == level &&
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
ucmp->CompareWithoutTimestamp(smallest_user_key,
c->GetLargestUserKey()) <= 0 &&
ucmp->CompareWithoutTimestamp(largest_user_key,
c->GetSmallestUserKey()) >= 0) {
// Overlap
return true;
}
if (c->SupportsPerKeyPlacement()) {
if (c->OverlapPenultimateLevelOutputRange(smallest_user_key,
largest_user_key)) {
return true;
}
}
}
// Did not overlap with any running compaction in level `level`
return false;
}
bool CompactionPicker::FilesRangeOverlapWithCompaction(
const std::vector<CompactionInputFiles>& inputs, int level,
int penultimate_level) const {
bool is_empty = true;
for (auto& in : inputs) {
if (!in.empty()) {
is_empty = false;
break;
}
}
if (is_empty) {
// No files in inputs
return false;
}
// TODO: Intra L0 compactions can have the ranges overlapped, but the input
// files cannot be overlapped in the order of L0 files.
InternalKey smallest, largest;
GetRange(inputs, &smallest, &largest, Compaction::kInvalidLevel);
if (penultimate_level != Compaction::kInvalidLevel) {
if (ioptions_.compaction_style == kCompactionStyleUniversal) {
if (RangeOverlapWithCompaction(smallest.user_key(), largest.user_key(),
penultimate_level)) {
return true;
}
} else {
InternalKey penultimate_smallest, penultimate_largest;
GetRange(inputs, &penultimate_smallest, &penultimate_largest, level);
if (RangeOverlapWithCompaction(penultimate_smallest.user_key(),
penultimate_largest.user_key(),
penultimate_level)) {
return true;
}
}
}
return RangeOverlapWithCompaction(smallest.user_key(), largest.user_key(),
level);
}
// Returns true if any one of specified files are being compacted
bool CompactionPicker::AreFilesInCompaction(
const std::vector<FileMetaData*>& files) {
for (size_t i = 0; i < files.size(); i++) {
if (files[i]->being_compacted) {
return true;
}
}
return false;
}
Compaction* CompactionPicker::CompactFiles(
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
const CompactionOptions& compact_options,
const std::vector<CompactionInputFiles>& input_files, int output_level,
VersionStorageInfo* vstorage, const MutableCFOptions& mutable_cf_options,
const MutableDBOptions& mutable_db_options, uint32_t output_path_id) {
#ifndef NDEBUG
assert(input_files.size());
// This compaction output should not overlap with a running compaction as
// `SanitizeCompactionInputFiles` should've checked earlier and db mutex
// shouldn't have been released since.
int start_level = Compaction::kInvalidLevel;
for (const auto& in : input_files) {
// input_files should already be sorted by level
if (!in.empty()) {
start_level = in.level;
break;
}
}
assert(output_level == 0 ||
!FilesRangeOverlapWithCompaction(
input_files, output_level,
Compaction::EvaluatePenultimateLevel(vstorage, ioptions_,
start_level, output_level)));
#endif /* !NDEBUG */
CompressionType compression_type;
if (compact_options.compression == kDisableCompressionOption) {
int base_level;
if (ioptions_.compaction_style == kCompactionStyleLevel) {
base_level = vstorage->base_level();
} else {
base_level = 1;
}
compression_type = GetCompressionType(vstorage, mutable_cf_options,
output_level, base_level);
} else {
// TODO(ajkr): `CompactionOptions` offers configurable `CompressionType`
// without configurable `CompressionOptions`, which is inconsistent.
compression_type = compact_options.compression;
}
auto c = new Compaction(
vstorage, ioptions_, mutable_cf_options, mutable_db_options, input_files,
output_level, compact_options.output_file_size_limit,
mutable_cf_options.max_compaction_bytes, output_path_id, compression_type,
GetCompressionOptions(mutable_cf_options, vstorage, output_level),
Temperature::kUnknown, compact_options.max_subcompactions,
/* grandparents */ {}, true);
RegisterCompaction(c);
return c;
}
Status CompactionPicker::GetCompactionInputsFromFileNumbers(
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
std::vector<CompactionInputFiles>* input_files,
std::unordered_set<uint64_t>* input_set, const VersionStorageInfo* vstorage,
const CompactionOptions& /*compact_options*/) const {
if (input_set->size() == 0U) {
return Status::InvalidArgument(
"Compaction must include at least one file.");
}
assert(input_files);
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
std::vector<CompactionInputFiles> matched_input_files;
matched_input_files.resize(vstorage->num_levels());
int first_non_empty_level = -1;
int last_non_empty_level = -1;
// TODO(yhchiang): use a lazy-initialized mapping from
// file_number to FileMetaData in Version.
for (int level = 0; level < vstorage->num_levels(); ++level) {
for (auto file : vstorage->LevelFiles(level)) {
auto iter = input_set->find(file->fd.GetNumber());
if (iter != input_set->end()) {
matched_input_files[level].files.push_back(file);
input_set->erase(iter);
last_non_empty_level = level;
if (first_non_empty_level == -1) {
first_non_empty_level = level;
}
}
}
}
if (!input_set->empty()) {
std::string message(
"Cannot find matched SST files for the following file numbers:");
for (auto fn : *input_set) {
message += " ";
message += std::to_string(fn);
}
return Status::InvalidArgument(message);
}
for (int level = first_non_empty_level; level <= last_non_empty_level;
++level) {
matched_input_files[level].level = level;
input_files->emplace_back(std::move(matched_input_files[level]));
}
return Status::OK();
}
// Returns true if any one of the parent files are being compacted
bool CompactionPicker::IsRangeInCompaction(VersionStorageInfo* vstorage,
const InternalKey* smallest,
const InternalKey* largest,
int level, int* level_index) {
std::vector<FileMetaData*> inputs;
assert(level < NumberLevels());
vstorage->GetOverlappingInputs(level, smallest, largest, &inputs,
level_index ? *level_index : 0, level_index);
return AreFilesInCompaction(inputs);
}
// Populates the set of inputs of all other levels that overlap with the
// start level.
// Now we assume all levels except start level and output level are empty.
// Will also attempt to expand "start level" if that doesn't expand
// "output level" or cause "level" to include a file for compaction that has an
// overlapping user-key with another file.
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
// REQUIRES: input_level and output_level are different
// REQUIRES: inputs->empty() == false
// Returns false if files on parent level are currently in compaction, which
// means that we can't compact them
bool CompactionPicker::SetupOtherInputs(
const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
VersionStorageInfo* vstorage, CompactionInputFiles* inputs,
CompactionInputFiles* output_level_inputs, int* parent_index,
Support subcmpct using reserved resources for round-robin priority (#10341) Summary: Earlier implementation of round-robin priority can only pick one file at a time and disallows parallel compactions within the same level. In this PR, round-robin compaction policy will expand towards more input files with respecting some additional constraints, which are summarized as follows: * Constraint 1: We can only pick consecutive files - Constraint 1a: When a file is being compacted (or some input files are being compacted after expanding), we cannot choose it and have to stop choosing more files - Constraint 1b: When we reach the last file (with the largest keys), we cannot choose more files (the next file will be the first one with small keys) * Constraint 2: We should ensure the total compaction bytes (including the overlapped files from the next level) is no more than `mutable_cf_options_.max_compaction_bytes` * Constraint 3: We try our best to pick as many files as possible so that the post-compaction level size can be just less than `MaxBytesForLevel(start_level_)` * Constraint 4: If trivial move is allowed, we reuse the logic of `TryNonL0TrivialMove()` instead of expanding files with Constraint 3 More details can be found in `LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion()`. The above optimization accelerates the process of moving the compaction cursor, in which the write-amp can be further reduced. While a large compaction may lead to high write stall, we break this large compaction into several subcompactions **regardless of** the `max_subcompactions` limit. The number of subcompactions for round-robin compaction priority is determined through the following steps: * Step 1: Initialized against `max_output_file_limit`, the number of input files in the start level, and also the range size limit `ranges.size()` * Step 2: Call `AcquireSubcompactionResources()`when max subcompactions is not sufficient, but we may or may not obtain desired resources, additional number of resources is stored in `extra_num_subcompaction_threads_reserved_`). Subcompaction limit is changed and update `num_planned_subcompactions` with `GetSubcompactionLimit()` * Step 3: Call `ShrinkSubcompactionResources()` to ensure extra resources can be released (extra resources may exist for round-robin compaction when the number of actual number of subcompactions is less than the number of planned subcompactions) More details can be found in `CompactionJob::AcquireSubcompactionResources()`,`CompactionJob::ShrinkSubcompactionResources()`, and `CompactionJob::ReleaseSubcompactionResources()`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10341 Test Plan: Add `CompactionPriMultipleFilesRoundRobin[1-3]` unit test in `compaction_picker_test.cc` and `RoundRobinSubcompactionsAgainstResources.SubcompactionsUsingResources/[0-4]`, `RoundRobinSubcompactionsAgainstPressureToken.PressureTokenTest/[0-1]` in `db_compaction_test.cc` Reviewed By: ajkr, hx235 Differential Revision: D37792644 Pulled By: littlepig2013 fbshipit-source-id: 7fecb7c4ffd97b34bbf6e3b760b2c35a772a0657
2 years ago
int base_index, bool only_expand_towards_right) {
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
assert(!inputs->empty());
assert(output_level_inputs->empty());
const int input_level = inputs->level;
const int output_level = output_level_inputs->level;
single-file bottom-level compaction when snapshot released Summary: When snapshots are held for a long time, files may reach the bottom level containing overwritten/deleted keys. We previously had no mechanism to trigger compaction on such files. This particularly impacted DBs that write to different parts of the keyspace over time, as such files would never be naturally compacted due to second-last level files moving down. This PR introduces a mechanism for bottommost files to be recompacted upon releasing all snapshots that prevent them from dropping their deleted/overwritten keys. - Changed `CompactionPicker` to compact files in `BottommostFilesMarkedForCompaction()`. These are the last choice when picking. Each file will be compacted alone and output to the same level in which it originated. The goal of this type of compaction is to rewrite the data excluding deleted/overwritten keys. - Changed `ReleaseSnapshot()` to recompute the bottom files marked for compaction when the oldest existing snapshot changes, and schedule a compaction if needed. We cache the value that oldest existing snapshot needs to exceed in order for another file to be marked in `bottommost_files_mark_threshold_`, which allows us to avoid recomputing marked files for most snapshot releases. - Changed `VersionStorageInfo` to track the list of bottommost files, which is recomputed every time the version changes by `UpdateBottommostFiles()`. The list of marked bottommost files is first computed in `ComputeBottommostFilesMarkedForCompaction()` when the version changes, but may also be recomputed when `ReleaseSnapshot()` is called. - Extracted core logic of `Compaction::IsBottommostLevel()` into `VersionStorageInfo::RangeMightExistAfterSortedRun()` since logic to check whether a file is bottommost is now necessary outside of compaction. Closes https://github.com/facebook/rocksdb/pull/3009 Differential Revision: D6062044 Pulled By: ajkr fbshipit-source-id: 123d201cf140715a7d5928e8b3cb4f9cd9f7ad21
7 years ago
if (input_level == output_level) {
// no possibility of conflict
return true;
}
// For now, we only support merging two levels, start level and output level.
// We need to assert other levels are empty.
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
for (int l = input_level + 1; l < output_level; l++) {
assert(vstorage->NumLevelFiles(l) == 0);
}
InternalKey smallest, largest;
// Get the range one last time.
GetRange(*inputs, &smallest, &largest);
// Populate the set of next-level files (inputs_GetOutputLevelInputs()) to
// include in compaction
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
vstorage->GetOverlappingInputs(output_level, &smallest, &largest,
&output_level_inputs->files, *parent_index,
parent_index);
if (AreFilesInCompaction(output_level_inputs->files)) {
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
return false;
}
if (!output_level_inputs->empty()) {
if (!ExpandInputsToCleanCut(cf_name, vstorage, output_level_inputs)) {
return false;
}
}
// See if we can further grow the number of inputs in "level" without
// changing the number of "level+1" files we pick up. We also choose NOT
// to expand if this would cause "level" to include some entries for some
// user key, while excluding other entries for the same user key. This
// can happen when one user key spans multiple files.
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
if (!output_level_inputs->empty()) {
const uint64_t limit = mutable_cf_options.max_compaction_bytes;
const uint64_t output_level_inputs_size =
TotalFileSize(output_level_inputs->files);
const uint64_t inputs_size = TotalFileSize(inputs->files);
bool expand_inputs = false;
CompactionInputFiles expanded_inputs;
expanded_inputs.level = input_level;
// Get closed interval of output level
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
InternalKey all_start, all_limit;
GetRange(*inputs, *output_level_inputs, &all_start, &all_limit);
bool try_overlapping_inputs = true;
Support subcmpct using reserved resources for round-robin priority (#10341) Summary: Earlier implementation of round-robin priority can only pick one file at a time and disallows parallel compactions within the same level. In this PR, round-robin compaction policy will expand towards more input files with respecting some additional constraints, which are summarized as follows: * Constraint 1: We can only pick consecutive files - Constraint 1a: When a file is being compacted (or some input files are being compacted after expanding), we cannot choose it and have to stop choosing more files - Constraint 1b: When we reach the last file (with the largest keys), we cannot choose more files (the next file will be the first one with small keys) * Constraint 2: We should ensure the total compaction bytes (including the overlapped files from the next level) is no more than `mutable_cf_options_.max_compaction_bytes` * Constraint 3: We try our best to pick as many files as possible so that the post-compaction level size can be just less than `MaxBytesForLevel(start_level_)` * Constraint 4: If trivial move is allowed, we reuse the logic of `TryNonL0TrivialMove()` instead of expanding files with Constraint 3 More details can be found in `LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion()`. The above optimization accelerates the process of moving the compaction cursor, in which the write-amp can be further reduced. While a large compaction may lead to high write stall, we break this large compaction into several subcompactions **regardless of** the `max_subcompactions` limit. The number of subcompactions for round-robin compaction priority is determined through the following steps: * Step 1: Initialized against `max_output_file_limit`, the number of input files in the start level, and also the range size limit `ranges.size()` * Step 2: Call `AcquireSubcompactionResources()`when max subcompactions is not sufficient, but we may or may not obtain desired resources, additional number of resources is stored in `extra_num_subcompaction_threads_reserved_`). Subcompaction limit is changed and update `num_planned_subcompactions` with `GetSubcompactionLimit()` * Step 3: Call `ShrinkSubcompactionResources()` to ensure extra resources can be released (extra resources may exist for round-robin compaction when the number of actual number of subcompactions is less than the number of planned subcompactions) More details can be found in `CompactionJob::AcquireSubcompactionResources()`,`CompactionJob::ShrinkSubcompactionResources()`, and `CompactionJob::ReleaseSubcompactionResources()`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10341 Test Plan: Add `CompactionPriMultipleFilesRoundRobin[1-3]` unit test in `compaction_picker_test.cc` and `RoundRobinSubcompactionsAgainstResources.SubcompactionsUsingResources/[0-4]`, `RoundRobinSubcompactionsAgainstPressureToken.PressureTokenTest/[0-1]` in `db_compaction_test.cc` Reviewed By: ajkr, hx235 Differential Revision: D37792644 Pulled By: littlepig2013 fbshipit-source-id: 7fecb7c4ffd97b34bbf6e3b760b2c35a772a0657
2 years ago
if (only_expand_towards_right) {
// Round-robin compaction only allows expansion towards the larger side.
vstorage->GetOverlappingInputs(input_level, &smallest, &all_limit,
&expanded_inputs.files, base_index,
nullptr);
} else {
vstorage->GetOverlappingInputs(input_level, &all_start, &all_limit,
&expanded_inputs.files, base_index,
nullptr);
}
uint64_t expanded_inputs_size = TotalFileSize(expanded_inputs.files);
if (!ExpandInputsToCleanCut(cf_name, vstorage, &expanded_inputs)) {
try_overlapping_inputs = false;
}
if (try_overlapping_inputs && expanded_inputs.size() > inputs->size() &&
Ignore max_compaction_bytes for compaction input that are within output key-range (#10835) Summary: When picking compaction input files, we sometimes stop picking a file that is fully included in the output key-range due to hitting max_compaction_bytes. Including these input files can potentially reduce WA at the expense of larger compactions. Larger compaction should be fine as files from input level are usually 10X smaller than files from output level. This PR adds a mutable CF option `ignore_max_compaction_bytes_for_input` that is enabled by default. We can remove this option once we are sure it is safe. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10835 Test Plan: - CI, a unit test on max_compaction_bytes fails before turning this flag off. - Benchmark does not show much difference in WA: `./db_bench --benchmarks=fillrandom,waitforcompaction,stats,levelstats -max_background_jobs=12 -num=2000000000 -target_file_size_base=33554432 --write_buffer_size=33554432` ``` main: ** Compaction Stats [default] ** Level Files Size Score Read(GB) Rn(GB) Rnp1(GB) Write(GB) Wnew(GB) Moved(GB) W-Amp Rd(MB/s) Wr(MB/s) Comp(sec) CompMergeCPU(sec) Comp(cnt) Avg(sec) KeyIn KeyDrop Rblob(GB) Wblob(GB) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ L0 3/0 91.59 MB 0.8 70.9 0.0 70.9 200.8 129.9 0.0 1.5 25.2 71.2 2886.55 2463.45 9725 0.297 1093M 254K 0.0 0.0 L1 9/0 248.03 MB 1.0 392.0 129.8 262.2 391.7 129.5 0.0 3.0 69.0 68.9 5821.71 5536.90 804 7.241 6029M 5814K 0.0 0.0 L2 87/0 2.50 GB 1.0 537.0 128.5 408.5 533.8 125.2 0.7 4.2 69.5 69.1 7912.24 7323.70 4417 1.791 8299M 36M 0.0 0.0 L3 836/0 24.99 GB 1.0 616.9 118.3 498.7 594.5 95.8 5.2 5.0 66.9 64.5 9442.38 8490.28 4204 2.246 9749M 306M 0.0 0.0 L4 2355/0 62.95 GB 0.3 67.3 37.1 30.2 54.2 24.0 38.9 1.5 72.2 58.2 954.37 821.18 917 1.041 1076M 173M 0.0 0.0 Sum 3290/0 90.77 GB 0.0 1684.2 413.7 1270.5 1775.0 504.5 44.9 13.7 63.8 67.3 27017.25 24635.52 20067 1.346 26G 522M 0.0 0.0 Cumulative compaction: 1774.96 GB write, 154.29 MB/s write, 1684.19 GB read, 146.40 MB/s read, 27017.3 seconds This PR: ** Compaction Stats [default] ** Level Files Size Score Read(GB) Rn(GB) Rnp1(GB) Write(GB) Wnew(GB) Moved(GB) W-Amp Rd(MB/s) Wr(MB/s) Comp(sec) CompMergeCPU(sec) Comp(cnt) Avg(sec) KeyIn KeyDrop Rblob(GB) Wblob(GB) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ L0 3/0 45.71 MB 0.8 72.9 0.0 72.9 202.8 129.9 0.0 1.6 25.4 70.7 2938.16 2510.36 9741 0.302 1124M 265K 0.0 0.0 L1 8/0 234.54 MB 0.9 384.5 129.8 254.7 384.2 129.6 0.0 3.0 69.0 68.9 5708.08 5424.43 791 7.216 5913M 5753K 0.0 0.0 L2 84/0 2.47 GB 1.0 543.1 128.6 414.5 539.9 125.4 0.7 4.2 69.6 69.2 7989.31 7403.13 4418 1.808 8393M 36M 0.0 0.0 L3 839/0 24.96 GB 1.0 615.6 118.4 497.2 593.2 96.0 5.1 5.0 66.6 64.1 9471.23 8489.31 4193 2.259 9726M 306M 0.0 0.0 L4 2360/0 63.04 GB 0.3 67.6 37.3 30.3 54.4 24.1 38.9 1.5 71.5 57.6 967.30 827.99 907 1.066 1080M 173M 0.0 0.0 Sum 3294/0 90.75 GB 0.0 1683.8 414.2 1269.6 1774.5 504.9 44.8 13.7 63.7 67.1 27074.08 24655.22 20050 1.350 26G 522M 0.0 0.0 Cumulative compaction: 1774.52 GB write, 157.09 MB/s write, 1683.77 GB read, 149.06 MB/s read, 27074.1 seconds ``` Reviewed By: ajkr Differential Revision: D40518319 Pulled By: cbi42 fbshipit-source-id: f4ea614bc0ebefe007ffaf05bb9aec9a8ca25b60
2 years ago
(mutable_cf_options.ignore_max_compaction_bytes_for_input ||
output_level_inputs_size + expanded_inputs_size < limit) &&
!AreFilesInCompaction(expanded_inputs.files)) {
InternalKey new_start, new_limit;
GetRange(expanded_inputs, &new_start, &new_limit);
CompactionInputFiles expanded_output_level_inputs;
expanded_output_level_inputs.level = output_level;
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
vstorage->GetOverlappingInputs(output_level, &new_start, &new_limit,
&expanded_output_level_inputs.files,
*parent_index, parent_index);
assert(!expanded_output_level_inputs.empty());
if (!AreFilesInCompaction(expanded_output_level_inputs.files) &&
ExpandInputsToCleanCut(cf_name, vstorage,
&expanded_output_level_inputs) &&
expanded_output_level_inputs.size() == output_level_inputs->size()) {
expand_inputs = true;
}
}
if (!expand_inputs) {
vstorage->GetCleanInputsWithinInterval(input_level, &all_start,
&all_limit, &expanded_inputs.files,
base_index, nullptr);
expanded_inputs_size = TotalFileSize(expanded_inputs.files);
if (expanded_inputs.size() > inputs->size() &&
Ignore max_compaction_bytes for compaction input that are within output key-range (#10835) Summary: When picking compaction input files, we sometimes stop picking a file that is fully included in the output key-range due to hitting max_compaction_bytes. Including these input files can potentially reduce WA at the expense of larger compactions. Larger compaction should be fine as files from input level are usually 10X smaller than files from output level. This PR adds a mutable CF option `ignore_max_compaction_bytes_for_input` that is enabled by default. We can remove this option once we are sure it is safe. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10835 Test Plan: - CI, a unit test on max_compaction_bytes fails before turning this flag off. - Benchmark does not show much difference in WA: `./db_bench --benchmarks=fillrandom,waitforcompaction,stats,levelstats -max_background_jobs=12 -num=2000000000 -target_file_size_base=33554432 --write_buffer_size=33554432` ``` main: ** Compaction Stats [default] ** Level Files Size Score Read(GB) Rn(GB) Rnp1(GB) Write(GB) Wnew(GB) Moved(GB) W-Amp Rd(MB/s) Wr(MB/s) Comp(sec) CompMergeCPU(sec) Comp(cnt) Avg(sec) KeyIn KeyDrop Rblob(GB) Wblob(GB) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ L0 3/0 91.59 MB 0.8 70.9 0.0 70.9 200.8 129.9 0.0 1.5 25.2 71.2 2886.55 2463.45 9725 0.297 1093M 254K 0.0 0.0 L1 9/0 248.03 MB 1.0 392.0 129.8 262.2 391.7 129.5 0.0 3.0 69.0 68.9 5821.71 5536.90 804 7.241 6029M 5814K 0.0 0.0 L2 87/0 2.50 GB 1.0 537.0 128.5 408.5 533.8 125.2 0.7 4.2 69.5 69.1 7912.24 7323.70 4417 1.791 8299M 36M 0.0 0.0 L3 836/0 24.99 GB 1.0 616.9 118.3 498.7 594.5 95.8 5.2 5.0 66.9 64.5 9442.38 8490.28 4204 2.246 9749M 306M 0.0 0.0 L4 2355/0 62.95 GB 0.3 67.3 37.1 30.2 54.2 24.0 38.9 1.5 72.2 58.2 954.37 821.18 917 1.041 1076M 173M 0.0 0.0 Sum 3290/0 90.77 GB 0.0 1684.2 413.7 1270.5 1775.0 504.5 44.9 13.7 63.8 67.3 27017.25 24635.52 20067 1.346 26G 522M 0.0 0.0 Cumulative compaction: 1774.96 GB write, 154.29 MB/s write, 1684.19 GB read, 146.40 MB/s read, 27017.3 seconds This PR: ** Compaction Stats [default] ** Level Files Size Score Read(GB) Rn(GB) Rnp1(GB) Write(GB) Wnew(GB) Moved(GB) W-Amp Rd(MB/s) Wr(MB/s) Comp(sec) CompMergeCPU(sec) Comp(cnt) Avg(sec) KeyIn KeyDrop Rblob(GB) Wblob(GB) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ L0 3/0 45.71 MB 0.8 72.9 0.0 72.9 202.8 129.9 0.0 1.6 25.4 70.7 2938.16 2510.36 9741 0.302 1124M 265K 0.0 0.0 L1 8/0 234.54 MB 0.9 384.5 129.8 254.7 384.2 129.6 0.0 3.0 69.0 68.9 5708.08 5424.43 791 7.216 5913M 5753K 0.0 0.0 L2 84/0 2.47 GB 1.0 543.1 128.6 414.5 539.9 125.4 0.7 4.2 69.6 69.2 7989.31 7403.13 4418 1.808 8393M 36M 0.0 0.0 L3 839/0 24.96 GB 1.0 615.6 118.4 497.2 593.2 96.0 5.1 5.0 66.6 64.1 9471.23 8489.31 4193 2.259 9726M 306M 0.0 0.0 L4 2360/0 63.04 GB 0.3 67.6 37.3 30.3 54.4 24.1 38.9 1.5 71.5 57.6 967.30 827.99 907 1.066 1080M 173M 0.0 0.0 Sum 3294/0 90.75 GB 0.0 1683.8 414.2 1269.6 1774.5 504.9 44.8 13.7 63.7 67.1 27074.08 24655.22 20050 1.350 26G 522M 0.0 0.0 Cumulative compaction: 1774.52 GB write, 157.09 MB/s write, 1683.77 GB read, 149.06 MB/s read, 27074.1 seconds ``` Reviewed By: ajkr Differential Revision: D40518319 Pulled By: cbi42 fbshipit-source-id: f4ea614bc0ebefe007ffaf05bb9aec9a8ca25b60
2 years ago
(mutable_cf_options.ignore_max_compaction_bytes_for_input ||
output_level_inputs_size + expanded_inputs_size < limit) &&
!AreFilesInCompaction(expanded_inputs.files)) {
expand_inputs = true;
}
}
if (expand_inputs) {
ROCKS_LOG_INFO(ioptions_.logger,
"[%s] Expanding@%d %" ROCKSDB_PRIszt "+%" ROCKSDB_PRIszt
"(%" PRIu64 "+%" PRIu64 " bytes) to %" ROCKSDB_PRIszt
"+%" ROCKSDB_PRIszt " (%" PRIu64 "+%" PRIu64 " bytes)\n",
cf_name.c_str(), input_level, inputs->size(),
output_level_inputs->size(), inputs_size,
output_level_inputs_size, expanded_inputs.size(),
output_level_inputs->size(), expanded_inputs_size,
output_level_inputs_size);
inputs->files = expanded_inputs.files;
}
} else {
// Likely to be trivial move. Expand files if they are still trivial moves,
// but limit to mutable_cf_options.max_compaction_bytes or 8 files so that
// we don't create too much compaction pressure for the next level.
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
return true;
}
void CompactionPicker::GetGrandparents(
VersionStorageInfo* vstorage, const CompactionInputFiles& inputs,
const CompactionInputFiles& output_level_inputs,
std::vector<FileMetaData*>* grandparents) {
InternalKey start, limit;
GetRange(inputs, output_level_inputs, &start, &limit);
// Compute the set of grandparent files that overlap this compaction
// (parent == level+1; grandparent == level+2 or the first
// level after that has overlapping files)
for (int level = output_level_inputs.level + 1; level < NumberLevels();
level++) {
vstorage->GetOverlappingInputs(level, &start, &limit, grandparents);
if (!grandparents->empty()) {
break;
}
}
}
Compaction* CompactionPicker::CompactRange(
const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
int input_level, int output_level,
const CompactRangeOptions& compact_range_options, const InternalKey* begin,
const InternalKey* end, InternalKey** compaction_end, bool* manual_conflict,
uint64_t max_file_num_to_ignore, const std::string& trim_ts) {
// CompactionPickerFIFO has its own implementation of compact range
assert(ioptions_.compaction_style != kCompactionStyleFIFO);
if (input_level == ColumnFamilyData::kCompactAllLevels) {
assert(ioptions_.compaction_style == kCompactionStyleUniversal);
// Universal compaction with more than one level always compacts all the
// files together to the last level.
assert(vstorage->num_levels() > 1);
// DBImpl::CompactRange() set output level to be the last level
if (ioptions_.allow_ingest_behind) {
assert(output_level == vstorage->num_levels() - 2);
} else {
assert(output_level == vstorage->num_levels() - 1);
}
// DBImpl::RunManualCompaction will make full range for universal compaction
assert(begin == nullptr);
assert(end == nullptr);
*compaction_end = nullptr;
int start_level = 0;
for (; start_level < vstorage->num_levels() &&
vstorage->NumLevelFiles(start_level) == 0;
start_level++) {
}
if (start_level == vstorage->num_levels()) {
return nullptr;
}
if ((start_level == 0) && (!level0_compactions_in_progress_.empty())) {
*manual_conflict = true;
// Only one level 0 compaction allowed
return nullptr;
}
std::vector<CompactionInputFiles> inputs(vstorage->num_levels() -
start_level);
for (int level = start_level; level < vstorage->num_levels(); level++) {
inputs[level - start_level].level = level;
auto& files = inputs[level - start_level].files;
for (FileMetaData* f : vstorage->LevelFiles(level)) {
files.push_back(f);
}
if (AreFilesInCompaction(files)) {
*manual_conflict = true;
return nullptr;
}
}
// 2 non-exclusive manual compactions could run at the same time producing
// overlaping outputs in the same level.
if (FilesRangeOverlapWithCompaction(
inputs, output_level,
Compaction::EvaluatePenultimateLevel(vstorage, ioptions_,
start_level, output_level))) {
// This compaction output could potentially conflict with the output
// of a currently running compaction, we cannot run it.
*manual_conflict = true;
return nullptr;
}
Compaction* c = new Compaction(
vstorage, ioptions_, mutable_cf_options, mutable_db_options,
std::move(inputs), output_level,
MaxFileSizeForLevel(mutable_cf_options, output_level,
ioptions_.compaction_style),
/* max_compaction_bytes */ LLONG_MAX,
compact_range_options.target_path_id,
GetCompressionType(vstorage, mutable_cf_options, output_level, 1),
GetCompressionOptions(mutable_cf_options, vstorage, output_level),
Temperature::kUnknown, compact_range_options.max_subcompactions,
/* grandparents */ {}, /* is manual */ true, trim_ts, /* score */ -1,
/* deletion_compaction */ false, /* l0_files_might_overlap */ true,
CompactionReason::kUnknown,
compact_range_options.blob_garbage_collection_policy,
compact_range_options.blob_garbage_collection_age_cutoff);
RegisterCompaction(c);
vstorage->ComputeCompactionScore(ioptions_, mutable_cf_options);
return c;
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
CompactionInputFiles inputs;
inputs.level = input_level;
bool covering_the_whole_range = true;
// All files are 'overlapping' in universal style compaction.
// We have to compact the entire range in one shot.
if (ioptions_.compaction_style == kCompactionStyleUniversal) {
begin = nullptr;
end = nullptr;
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
vstorage->GetOverlappingInputs(input_level, begin, end, &inputs.files);
if (inputs.empty()) {
return nullptr;
}
if ((input_level == 0) && (!level0_compactions_in_progress_.empty())) {
// Only one level 0 compaction allowed
TEST_SYNC_POINT("CompactionPicker::CompactRange:Conflict");
*manual_conflict = true;
return nullptr;
}
// Avoid compacting too much in one shot in case the range is large.
// But we cannot do this for level-0 since level-0 files can overlap
// and we must not pick one file and drop another older file if the
// two files overlap.
if (input_level > 0) {
const uint64_t limit = mutable_cf_options.max_compaction_bytes;
uint64_t input_level_total = 0;
int hint_index = -1;
InternalKey* smallest = nullptr;
InternalKey* largest = nullptr;
for (size_t i = 0; i + 1 < inputs.size(); ++i) {
if (!smallest) {
smallest = &inputs[i]->smallest;
}
largest = &inputs[i]->largest;
uint64_t input_file_size = inputs[i]->fd.GetFileSize();
uint64_t output_level_total = 0;
if (output_level < vstorage->num_non_empty_levels()) {
std::vector<FileMetaData*> files;
vstorage->GetOverlappingInputsRangeBinarySearch(
output_level, smallest, largest, &files, hint_index, &hint_index);
for (const auto& file : files) {
output_level_total += file->fd.GetFileSize();
}
}
input_level_total += input_file_size;
if (input_level_total + output_level_total >= limit) {
covering_the_whole_range = false;
// still include the current file, so the compaction could be larger
// than max_compaction_bytes, which is also to make sure the compaction
// can make progress even `max_compaction_bytes` is small (e.g. smaller
// than an SST file).
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
inputs.files.resize(i + 1);
break;
}
}
}
assert(compact_range_options.target_path_id <
static_cast<uint32_t>(ioptions_.cf_paths.size()));
// for BOTTOM LEVEL compaction only, use max_file_num_to_ignore to filter out
// files that are created during the current compaction.
if (compact_range_options.bottommost_level_compaction ==
BottommostLevelCompaction::kForceOptimized &&
max_file_num_to_ignore != std::numeric_limits<uint64_t>::max()) {
assert(input_level == output_level);
// inputs_shrunk holds a continuous subset of input files which were all
// created before the current manual compaction
std::vector<FileMetaData*> inputs_shrunk;
size_t skip_input_index = inputs.size();
for (size_t i = 0; i < inputs.size(); ++i) {
if (inputs[i]->fd.GetNumber() < max_file_num_to_ignore) {
inputs_shrunk.push_back(inputs[i]);
} else if (!inputs_shrunk.empty()) {
// inputs[i] was created during the current manual compaction and
// need to be skipped
skip_input_index = i;
break;
}
}
if (inputs_shrunk.empty()) {
return nullptr;
}
if (inputs.size() != inputs_shrunk.size()) {
inputs.files.swap(inputs_shrunk);
}
// set covering_the_whole_range to false if there is any file that need to
// be compacted in the range of inputs[skip_input_index+1, inputs.size())
for (size_t i = skip_input_index + 1; i < inputs.size(); ++i) {
if (inputs[i]->fd.GetNumber() < max_file_num_to_ignore) {
covering_the_whole_range = false;
}
}
}
InternalKey key_storage;
InternalKey* next_smallest = &key_storage;
if (ExpandInputsToCleanCut(cf_name, vstorage, &inputs, &next_smallest) ==
false) {
// manual compaction is now multi-threaded, so it can
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
// happen that ExpandWhileOverlapping fails
// we handle it higher in RunManualCompaction
*manual_conflict = true;
return nullptr;
}
if (covering_the_whole_range || !next_smallest) {
*compaction_end = nullptr;
} else {
**compaction_end = *next_smallest;
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
CompactionInputFiles output_level_inputs;
if (output_level == ColumnFamilyData::kCompactToBaseLevel) {
assert(input_level == 0);
output_level = vstorage->base_level();
assert(output_level > 0);
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
output_level_inputs.level = output_level;
if (input_level != output_level) {
int parent_index = -1;
if (!SetupOtherInputs(cf_name, mutable_cf_options, vstorage, &inputs,
&output_level_inputs, &parent_index, -1)) {
// manual compaction is now multi-threaded, so it can
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
// happen that SetupOtherInputs fails
// we handle it higher in RunManualCompaction
*manual_conflict = true;
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
return nullptr;
}
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
std::vector<CompactionInputFiles> compaction_inputs({inputs});
if (!output_level_inputs.empty()) {
compaction_inputs.push_back(output_level_inputs);
}
for (size_t i = 0; i < compaction_inputs.size(); i++) {
if (AreFilesInCompaction(compaction_inputs[i].files)) {
*manual_conflict = true;
return nullptr;
}
}
// 2 non-exclusive manual compactions could run at the same time producing
// overlaping outputs in the same level.
if (FilesRangeOverlapWithCompaction(
compaction_inputs, output_level,
Compaction::EvaluatePenultimateLevel(vstorage, ioptions_, input_level,
output_level))) {
// This compaction output could potentially conflict with the output
// of a currently running compaction, we cannot run it.
*manual_conflict = true;
return nullptr;
}
Make Compaction class easier to use Summary: The goal of this diff is to make Compaction class easier to use. This should also make new compaction algorithms easier to write (like CompactFiles from @yhchiang and dynamic leveled and multi-leveled universal from @sdong). Here are couple of things demonstrating that Compaction class is hard to use: 1. we have two constructors of Compaction class 2. there's this thing called grandparents_, but it appears to only be setup for leveled compaction and not compactfiles 3. it's easy to introduce a subtle and dangerous bug like this: D36225 4. SetupBottomMostLevel() is hard to understand and it shouldn't be. See this comment: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction.cc#L236-L241. It also made it harder for @yhchiang to write CompactFiles, as evidenced by this: https://github.com/facebook/rocksdb/blob/afbafeaeaebfd27a0f3e992fee8e0c57d07658fa/db/compaction_picker.cc#L204-L210 The problem is that we create Compaction object, which holds a lot of state, and then pass it around to some functions. After those functions are done mutating, then we call couple of functions on Compaction object, like SetupBottommostLevel() and MarkFilesBeingCompacted(). It is very hard to see what's happening with all that Compaction's state while it's travelling across different functions. If you're writing a new PickCompaction() function you need to try really hard to understand what are all the functions you need to run on Compaction object and what state you need to setup. My proposed solution is to make important parts of Compaction immutable after construction. PickCompaction() should calculate compaction inputs and then pass them onto Compaction object once they are finalized. That makes it easy to create a new compaction -- just provide all the parameters to the constructor and you're done. No need to call confusing functions after you created your object. This diff doesn't fully achieve that goal, but it comes pretty close. Here are some of the changes: * have one Compaction constructor instead of two. * inputs_ is constant after construction * MarkFilesBeingCompacted() is now private to Compaction class and automatically called on construction/destruction. * SetupBottommostLevel() is gone. Compaction figures it out on its own based on the input. * CompactionPicker's functions are not passing around Compaction object anymore. They are only passing around the state that they need. Test Plan: make check make asan_check make valgrind_check Reviewers: rven, anthony, sdong, yhchiang Reviewed By: yhchiang Subscribers: sdong, yhchiang, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D36687
10 years ago
std::vector<FileMetaData*> grandparents;
GetGrandparents(vstorage, inputs, output_level_inputs, &grandparents);
Compaction* compaction = new Compaction(
vstorage, ioptions_, mutable_cf_options, mutable_db_options,
std::move(compaction_inputs), output_level,
MaxFileSizeForLevel(mutable_cf_options, output_level,
ioptions_.compaction_style, vstorage->base_level(),
ioptions_.level_compaction_dynamic_level_bytes),
mutable_cf_options.max_compaction_bytes,
compact_range_options.target_path_id,
GetCompressionType(vstorage, mutable_cf_options, output_level,
vstorage->base_level()),
GetCompressionOptions(mutable_cf_options, vstorage, output_level),
Temperature::kUnknown, compact_range_options.max_subcompactions,
std::move(grandparents), /* is manual */ true, trim_ts, /* score */ -1,
/* deletion_compaction */ false, /* l0_files_might_overlap */ true,
CompactionReason::kUnknown,
compact_range_options.blob_garbage_collection_policy,
compact_range_options.blob_garbage_collection_age_cutoff);
TEST_SYNC_POINT_CALLBACK("CompactionPicker::CompactRange:Return", compaction);
RegisterCompaction(compaction);
// Creating a compaction influences the compaction score because the score
// takes running compactions into account (by skipping files that are already
// being compacted). Since we just changed compaction score, we recalculate it
// here
vstorage->ComputeCompactionScore(ioptions_, mutable_cf_options);
return compaction;
}
#ifndef ROCKSDB_LITE
namespace {
// Test whether two files have overlapping key-ranges.
bool HaveOverlappingKeyRanges(const Comparator* c, const SstFileMetaData& a,
const SstFileMetaData& b) {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (c->CompareWithoutTimestamp(a.smallestkey, b.smallestkey) >= 0) {
if (c->CompareWithoutTimestamp(a.smallestkey, b.largestkey) <= 0) {
// b.smallestkey <= a.smallestkey <= b.largestkey
return true;
}
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
} else if (c->CompareWithoutTimestamp(a.largestkey, b.smallestkey) >= 0) {
// a.smallestkey < b.smallestkey <= a.largestkey
return true;
}
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (c->CompareWithoutTimestamp(a.largestkey, b.largestkey) <= 0) {
if (c->CompareWithoutTimestamp(a.largestkey, b.smallestkey) >= 0) {
// b.smallestkey <= a.largestkey <= b.largestkey
return true;
}
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
} else if (c->CompareWithoutTimestamp(a.smallestkey, b.largestkey) <= 0) {
// a.smallestkey <= b.largestkey < a.largestkey
return true;
}
return false;
}
} // namespace
Status CompactionPicker::SanitizeCompactionInputFilesForAllLevels(
std::unordered_set<uint64_t>* input_files,
const ColumnFamilyMetaData& cf_meta, const int output_level) const {
auto& levels = cf_meta.levels;
auto comparator = icmp_->user_comparator();
// TODO(yhchiang): add is_adjustable to CompactionOptions
// the smallest and largest key of the current compaction input
std::string smallestkey;
std::string largestkey;
// a flag for initializing smallest and largest key
bool is_first = false;
const int kNotFound = -1;
// For each level, it does the following things:
// 1. Find the first and the last compaction input files
// in the current level.
// 2. Include all files between the first and the last
// compaction input files.
// 3. Update the compaction key-range.
// 4. For all remaining levels, include files that have
// overlapping key-range with the compaction key-range.
for (int l = 0; l <= output_level; ++l) {
auto& current_files = levels[l].files;
int first_included = static_cast<int>(current_files.size());
int last_included = kNotFound;
// identify the first and the last compaction input files
// in the current level.
for (size_t f = 0; f < current_files.size(); ++f) {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
const uint64_t file_number = TableFileNameToNumber(current_files[f].name);
if (input_files->find(file_number) == input_files->end()) {
continue;
}
first_included = std::min(first_included, static_cast<int>(f));
last_included = std::max(last_included, static_cast<int>(f));
if (is_first == false) {
smallestkey = current_files[f].smallestkey;
largestkey = current_files[f].largestkey;
is_first = true;
}
}
if (last_included == kNotFound) {
continue;
}
if (l != 0) {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
// expand the compaction input of the current level if it
// has overlapping key-range with other non-compaction input
// files in the same level.
while (first_included > 0) {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (comparator->CompareWithoutTimestamp(
current_files[first_included - 1].largestkey,
current_files[first_included].smallestkey) < 0) {
break;
}
first_included--;
}
while (last_included < static_cast<int>(current_files.size()) - 1) {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (comparator->CompareWithoutTimestamp(
current_files[last_included + 1].smallestkey,
current_files[last_included].largestkey) > 0) {
break;
}
last_included++;
}
} else if (output_level > 0) {
last_included = static_cast<int>(current_files.size() - 1);
}
// include all files between the first and the last compaction input files.
for (int f = first_included; f <= last_included; ++f) {
if (current_files[f].being_compacted) {
return Status::Aborted("Necessary compaction input file " +
current_files[f].name +
" is currently being compacted.");
}
Sort L0 files by newly introduced epoch_num (#10922) Summary: **Context:** Sorting L0 files by `largest_seqno` has at least two inconvenience: - File ingestion and compaction involving ingested files can create files of overlapping seqno range with the existing files. `force_consistency_check=true` will catch such overlap seqno range even those harmless overlap. - For example, consider the following sequence of events ("key@n" indicates key at seqno "n") - insert k1@1 to memtable m1 - ingest file s1 with k2@2, ingest file s2 with k3@3 - insert k4@4 to m1 - compact files s1, s2 and result in new file s3 of seqno range [2, 3] - flush m1 and result in new file s4 of seqno range [1, 4]. And `force_consistency_check=true` will think s4 and s3 has file reordering corruption that might cause retuning an old value of k1 - However such caught corruption is a false positive since s1, s2 will not have overlapped keys with k1 or whatever inserted into m1 before ingest file s1 by the requirement of file ingestion (otherwise the m1 will be flushed first before any of the file ingestion completes). Therefore there in fact isn't any file reordering corruption. - Single delete can decrease a file's largest seqno and ordering by `largest_seqno` can introduce a wrong ordering hence file reordering corruption - For example, consider the following sequence of events ("key@n" indicates key at seqno "n", Credit to ajkr for this example) - an existing SST s1 contains only k1@1 - insert k1@2 to memtable m1 - ingest file s2 with k3@3, ingest file s3 with k4@4 - insert single delete k5@5 in m1 - flush m1 and result in new file s4 of seqno range [2, 5] - compact s1, s2, s3 and result in new file s5 of seqno range [1, 4] - compact s4 and result in new file s6 of seqno range [2] due to single delete - By the last step, we have file ordering by largest seqno (">" means "newer") : s5 > s6 while s6 contains a newer version of the k1's value (i.e, k1@2) than s5, which is a real reordering corruption. While this can be caught by `force_consistency_check=true`, there isn't a good way to prevent this from happening if ordering by `largest_seqno` Therefore, we are redesigning the sorting criteria of L0 files and avoid above inconvenience. Credit to ajkr , we now introduce `epoch_num` which describes the order of a file being flushed or ingested/imported (compaction output file will has the minimum `epoch_num` among input files'). This will avoid the above inconvenience in the following ways: - In the first case above, there will no longer be overlap seqno range check in `force_consistency_check=true` but `epoch_number` ordering check. This will result in file ordering s1 < s2 < s4 (pre-compaction) and s3 < s4 (post-compaction) which won't trigger false positive corruption. See test class `DBCompactionTestL0FilesMisorderCorruption*` for more. - In the second case above, this will result in file ordering s1 < s2 < s3 < s4 (pre-compacting s1, s2, s3), s5 < s4 (post-compacting s1, s2, s3), s5 < s6 (post-compacting s4), which are correct file ordering without causing any corruption. **Summary:** - Introduce `epoch_number` stored per `ColumnFamilyData` and sort CF's L0 files by their assigned `epoch_number` instead of `largest_seqno`. - `epoch_number` is increased and assigned upon `VersionEdit::AddFile()` for flush (or similarly for WriteLevel0TableForRecovery) and file ingestion (except for allow_behind_true, which will always get assigned as the `kReservedEpochNumberForFileIngestedBehind`) - Compaction output file is assigned with the minimum `epoch_number` among input files' - Refit level: reuse refitted file's epoch_number - Other paths needing `epoch_number` treatment: - Import column families: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo` - Repair: reuse file's epoch_number if exists. If not, assign one based on `NewestFirstBySeqNo`. - Assigning new epoch_number to a file and adding this file to LSM tree should be atomic. This is guaranteed by us assigning epoch_number right upon `VersionEdit::AddFile()` where this version edit will be apply to LSM tree shape right after by holding the db mutex (e.g, flush, file ingestion, import column family) or by there is only 1 ongoing edit per CF (e.g, WriteLevel0TableForRecovery, Repair). - Assigning the minimum input epoch number to compaction output file won't misorder L0 files (even through later `Refit(target_level=0)`). It's due to for every key "k" in the input range, a legit compaction will cover a continuous epoch number range of that key. As long as we assign the key "k" the minimum input epoch number, it won't become newer or older than the versions of this key that aren't included in this compaction hence no misorder. - Persist `epoch_number` of each file in manifest and recover `epoch_number` on db recovery - Backward compatibility with old db without `epoch_number` support is guaranteed by assigning `epoch_number` to recovered files by `NewestFirstBySeqno` order. See `VersionStorageInfo::RecoverEpochNumbers()` for more - Forward compatibility with manifest is guaranteed by flexibility of `NewFileCustomTag` - Replace `force_consistent_check` on L0 with `epoch_number` and remove false positive check like case 1 with `largest_seqno` above - Due to backward compatibility issue, we might encounter files with missing epoch number at the beginning of db recovery. We will still use old L0 sorting mechanism (`NewestFirstBySeqno`) to check/sort them till we infer their epoch number. See usages of `EpochNumberRequirement`. - Remove fix https://github.com/facebook/rocksdb/pull/5958#issue-511150930 and their outdated tests to file reordering corruption because such fix can be replaced by this PR. - Misc: - update existing tests with `epoch_number` so make check will pass - update https://github.com/facebook/rocksdb/pull/5958#issue-511150930 tests to verify corruption is fixed using `epoch_number` and cover universal/fifo compaction/CompactRange/CompactFile cases - assert db_mutex is held for a few places before calling ColumnFamilyData::NewEpochNumber() Pull Request resolved: https://github.com/facebook/rocksdb/pull/10922 Test Plan: - `make check` - New unit tests under `db/db_compaction_test.cc`, `db/db_test2.cc`, `db/version_builder_test.cc`, `db/repair_test.cc` - Updated tests (i.e, `DBCompactionTestL0FilesMisorderCorruption*`) under https://github.com/facebook/rocksdb/pull/5958#issue-511150930 - [Ongoing] Compatibility test: manually run https://github.com/ajkr/rocksdb/commit/36a5686ec012f35a4371e409aa85c404ca1c210d (with file ingestion off for running the `.orig` binary to prevent this bug affecting upgrade/downgrade formality checking) for 1 hour on `simple black/white box`, `cf_consistency/txn/enable_ts with whitebox + test_best_efforts_recovery with blackbox` - [Ongoing] normal db stress test - [Ongoing] db stress test with aggressive value https://github.com/facebook/rocksdb/pull/10761 Reviewed By: ajkr Differential Revision: D41063187 Pulled By: hx235 fbshipit-source-id: 826cb23455de7beaabe2d16c57682a82733a32a9
2 years ago
input_files->insert(TableFileNameToNumber(current_files[f].name));
}
// update smallest and largest key
if (l == 0) {
for (int f = first_included; f <= last_included; ++f) {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (comparator->CompareWithoutTimestamp(
smallestkey, current_files[f].smallestkey) > 0) {
smallestkey = current_files[f].smallestkey;
}
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (comparator->CompareWithoutTimestamp(
largestkey, current_files[f].largestkey) < 0) {
largestkey = current_files[f].largestkey;
}
}
} else {
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (comparator->CompareWithoutTimestamp(
smallestkey, current_files[first_included].smallestkey) > 0) {
smallestkey = current_files[first_included].smallestkey;
}
Fix overlapping check by excluding timestamp (#10615) Summary: With user-defined timestamp, checking overlapping should exclude timestamp part from key. This has already been done for range checking for files in sstableKeyCompare(), but not yet done when checking with concurrent compactions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/10615 Test Plan: (Will add more tests) make check (Repro seems easier with this commit sha: git checkout 78bbdef530bd36fa299d496bd1013cf39d8e203a) rm -rf /dev/shm/rocksdb/* && mkdir /dev/shm/rocksdb/rocksdb_crashtest_expected && ./db_stress --allow_data_in_errors=True --clear_column_family_one_in=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb//rocksdb_crashtest_blackbox --delpercent=5 --delrangepercent=0 --expected_values_dir=/dev/shm/rocksdb//rocksdb_crashtest_expected --iterpercent=0 --max_background_compactions=20 --max_bytes_for_level_base=10485760 --max_key=25000000 --max_write_batch_group_size_bytes=1048576 --nooverwritepercent=1 --ops_per_thread=1000000 --paranoid_file_checks=1 --partition_filters=0 --prefix_size=8 --prefixpercent=5 --readpercent=30 --reopen=0 --snapshot_hold_ops=100000 --subcompactions=1 --compaction_pri=3 --target_file_size_base=65536 --target_file_size_multiplier=2 --test_batches_snapshots=0 --test_cf_consistency=0 --use_multiget=1 --user_timestamp_size=8 --value_size_mult=32 --verify_checksum=1 --write_buffer_size=65536 --writepercent=60 -disable_wal=1 Reviewed By: akankshamahajan15 Differential Revision: D39146797 Pulled By: riversand963 fbshipit-source-id: 7fca800026ca6219220100b8b6cf84d907828163
2 years ago
if (comparator->CompareWithoutTimestamp(
largestkey, current_files[last_included].largestkey) < 0) {
largestkey = current_files[last_included].largestkey;
}
}
SstFileMetaData aggregated_file_meta;
aggregated_file_meta.smallestkey = smallestkey;
aggregated_file_meta.largestkey = largestkey;
// For all lower levels, include all overlapping files.
Fix CompactFiles by adding all necessary files Summary: The compact files API had a bug where some overlapping files are not added. These are files which overlap with files which were added to the compaction input files, but not to the original set of input files. This happens only when there are more than two levels involved in the compaction. An example will illustrate this better. Level 2 has 1 input file 1.sst which spans [20,30]. Level 3 has added file 2.sst which spans [10,25] Level 4 has file 3.sst which spans [35,40] and input file 4.sst which spans [46,50]. The existing code would not add 3.sst to the set of input_files because it only becomes an overlapping file in level 4 and it wasn't one in level 3. When installing the results of the compaction, 3.sst would overlap with output file from the compact files and result in the assertion in version_set.cc:1130 // Must not overlap assert(level <= 0 || level_files->empty() || internal_comparator_->Compare( (*level_files)[level_files->size() - 1]->largest, f->smallest) < 0); This change now adds overlapping files from the current level to the set of input files also so that we don't hit the assertion above. Test Plan: d=/tmp/j; rm -rf $d; seq 1000 | parallel --gnu --eta 'd=/tmp/j/d-{}; mkdir -p $d; TEST_TMPDIR=$d ./db_compaction_test --gtest_filter=*CompactilesOnLevel* --gtest_also_run_disabled_tests >& '$d'/log-{}' Reviewers: igor, yhchiang, sdong Reviewed By: yhchiang Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D43437
10 years ago
// We need to add overlapping files from the current level too because even
// if there no input_files in level l, we would still need to add files
// which overlap with the range containing the input_files in levels 0 to l
// Level 0 doesn't need to be handled this way because files are sorted by
// time and not by key
for (int m = std::max(l, 1); m <= output_level; ++m) {
for (auto& next_lv_file : levels[m].files) {
if (HaveOverlappingKeyRanges(comparator, aggregated_file_meta,
next_lv_file)) {
if (next_lv_file.being_compacted) {
return Status::Aborted(
"File " + next_lv_file.name +
" that has overlapping key range with one of the compaction "
" input file is currently being compacted.");
}
input_files->insert(TableFileNameToNumber(next_lv_file.name));
}
}
}
}
if (RangeOverlapWithCompaction(smallestkey, largestkey, output_level)) {
return Status::Aborted(
"A running compaction is writing to the same output level in an "
"overlapping key range");
}
return Status::OK();
}
Status CompactionPicker::SanitizeCompactionInputFiles(
std::unordered_set<uint64_t>* input_files,
const ColumnFamilyMetaData& cf_meta, const int output_level) const {
assert(static_cast<int>(cf_meta.levels.size()) - 1 ==
cf_meta.levels[cf_meta.levels.size() - 1].level);
if (output_level >= static_cast<int>(cf_meta.levels.size())) {
return Status::InvalidArgument(
"Output level for column family " + cf_meta.name +
" must between [0, " +
std::to_string(cf_meta.levels[cf_meta.levels.size() - 1].level) + "].");
}
if (output_level > MaxOutputLevel()) {
return Status::InvalidArgument(
"Exceed the maximum output level defined by "
"the current compaction algorithm --- " +
std::to_string(MaxOutputLevel()));
}
if (output_level < 0) {
return Status::InvalidArgument("Output level cannot be negative.");
}
if (input_files->size() == 0) {
return Status::InvalidArgument(
"A compaction must contain at least one file.");
}
Status s = SanitizeCompactionInputFilesForAllLevels(input_files, cf_meta,
output_level);
if (!s.ok()) {
return s;
}
// for all input files, check whether the file number matches
// any currently-existing files.
for (auto file_num : *input_files) {
bool found = false;
int input_file_level = -1;
for (const auto& level_meta : cf_meta.levels) {
for (const auto& file_meta : level_meta.files) {
if (file_num == TableFileNameToNumber(file_meta.name)) {
if (file_meta.being_compacted) {
return Status::Aborted("Specified compaction input file " +
MakeTableFileName("", file_num) +
" is already being compacted.");
}
found = true;
input_file_level = level_meta.level;
break;
}
}
if (found) {
break;
}
}
if (!found) {
return Status::InvalidArgument(
"Specified compaction input file " + MakeTableFileName("", file_num) +
" does not exist in column family " + cf_meta.name + ".");
}
if (input_file_level > output_level) {
return Status::InvalidArgument(
"Cannot compact file to up level, input file: " +
MakeTableFileName("", file_num) + " level " +
std::to_string(input_file_level) + " > output level " +
std::to_string(output_level));
}
}
return Status::OK();
}
#endif // !ROCKSDB_LITE
void CompactionPicker::RegisterCompaction(Compaction* c) {
if (c == nullptr) {
return;
}
assert(ioptions_.compaction_style != kCompactionStyleLevel ||
c->output_level() == 0 ||
!FilesRangeOverlapWithCompaction(*c->inputs(), c->output_level(),
c->GetPenultimateLevel()));
Add missing range conflict check between file ingestion and RefitLevel() (#10988) Summary: **Context:** File ingestion never checks whether the key range it acts on overlaps with an ongoing RefitLevel() (used in `CompactRange()` with `change_level=true`). That's because RefitLevel() doesn't register and make its key range known to file ingestion. Though it checks overlapping with other compactions by https://github.com/facebook/rocksdb/blob/7.8.fb/db/external_sst_file_ingestion_job.cc#L998. RefitLevel() (used in `CompactRange()` with `change_level=true`) doesn't check whether the key range it acts on overlaps with an ongoing file ingestion. That's because file ingestion does not register and make its key range known to other compactions. - Note that non-refitlevel-compaction (e.g, manual compaction w/o RefitLevel() or general compaction) also does not check key range overlap with ongoing file ingestion for the same reason. - But it's fine. Credited to cbi42's discovery, `WaitForIngestFile` was called by background and foreground compactions. They were introduced in https://github.com/facebook/rocksdb/commit/0f88160f67d36ea30e3aca3a3cef924c3a009be6, https://github.com/facebook/rocksdb/commit/5c64fb67d2fc198f1a73ff3ae543749a6a41f513 and https://github.com/facebook/rocksdb/commit/87dfc1d23e0e16ff73e15f63c6fa0fb3b3fc8c8c. - Regardless, this PR registers file ingestion like a compaction is a general approach that will also add range conflict check between file ingestion and non-refitlevel-compaction, though it has not been the issue motivated this PR. Above are bugs resulting in two bad consequences: - If file ingestion and RefitLevel() creates files in the same level, then range-overlapped files will be created at that level and caught as corruption by `force_consistency_checks=true` - If file ingestion and RefitLevel() creates file in different levels, then with one further compaction on the ingested file, it can result in two same keys both with seqno 0 in two different levels. Then with iterator's [optimization](https://github.com/facebook/rocksdb/blame/c62f3221698fd273b673d4f7e54eabb8329a4369/db/db_iter.cc#L342-L343) that assumes no two same keys both with seqno 0, it will either break this assertion in debug build or, even worst, return value of this same key for the key after it, which is the wrong value to return, in release build. Therefore we decide to introduce range conflict check for file ingestion and RefitLevel() inspired from the existing range conflict check among compactions. **Summary:** - Treat file ingestion job and RefitLevel() as `Compaction` of new compaction reasons: `CompactionReason::kExternalSstIngestion` and `CompactionReason::kRefitLevel` and register/unregister them. File ingestion is treated as compaction from L0 to different levels and RefitLevel() as compaction from source level to target level. - Check for `RangeOverlapWithCompaction` with other ongoing compactions, `RegisterCompaction()` on this "compaction" before changing the LSM state in `VersionStorageInfo`, and `UnregisterCompaction()` after changing. - Replace scattered fixes (https://github.com/facebook/rocksdb/commit/0f88160f67d36ea30e3aca3a3cef924c3a009be6, https://github.com/facebook/rocksdb/commit/5c64fb67d2fc198f1a73ff3ae543749a6a41f513 and https://github.com/facebook/rocksdb/commit/87dfc1d23e0e16ff73e15f63c6fa0fb3b3fc8c8c.) that prevents overlapping between file ingestion and non-refit-level compaction with this fix cuz those practices are easy to overlook. - Misc: logic cleanup, see PR comments Pull Request resolved: https://github.com/facebook/rocksdb/pull/10988 Test Plan: - New unit test `DBCompactionTestWithOngoingFileIngestionParam*` that failed pre-fix and passed afterwards. - Made compatible with existing tests, see PR comments - make check - [Ongoing] Stress test rehearsal with normal value and aggressive CI value https://github.com/facebook/rocksdb/pull/10761 Reviewed By: cbi42 Differential Revision: D41535685 Pulled By: hx235 fbshipit-source-id: 549833a577ba1496d20a870583d4caa737da1258
2 years ago
// CompactionReason::kExternalSstIngestion's start level is just a placeholder
// number without actual meaning as file ingestion technically does not have
// an input level like other compactions
if ((c->start_level() == 0 &&
c->compaction_reason() != CompactionReason::kExternalSstIngestion) ||
ioptions_.compaction_style == kCompactionStyleUniversal) {
level0_compactions_in_progress_.insert(c);
}
compactions_in_progress_.insert(c);
TEST_SYNC_POINT_CALLBACK("CompactionPicker::RegisterCompaction:Registered",
c);
}
void CompactionPicker::UnregisterCompaction(Compaction* c) {
if (c == nullptr) {
return;
}
if (c->start_level() == 0 ||
ioptions_.compaction_style == kCompactionStyleUniversal) {
level0_compactions_in_progress_.erase(c);
}
compactions_in_progress_.erase(c);
}
void CompactionPicker::PickFilesMarkedForCompaction(
const std::string& cf_name, VersionStorageInfo* vstorage, int* start_level,
int* output_level, CompactionInputFiles* start_level_inputs) {
if (vstorage->FilesMarkedForCompaction().empty()) {
return;
}
auto continuation = [&, cf_name](std::pair<int, FileMetaData*> level_file) {
// If it's being compacted it has nothing to do here.
// If this assert() fails that means that some function marked some
// files as being_compacted, but didn't call ComputeCompactionScore()
assert(!level_file.second->being_compacted);
*start_level = level_file.first;
*output_level =
(*start_level == 0) ? vstorage->base_level() : *start_level + 1;
if (*start_level == 0 && !level0_compactions_in_progress()->empty()) {
return false;
}
start_level_inputs->files = {level_file.second};
start_level_inputs->level = *start_level;
return ExpandInputsToCleanCut(cf_name, vstorage, start_level_inputs);
};
// take a chance on a random file first
Random64 rnd(/* seed */ reinterpret_cast<uint64_t>(vstorage));
size_t random_file_index = static_cast<size_t>(rnd.Uniform(
static_cast<uint64_t>(vstorage->FilesMarkedForCompaction().size())));
TEST_SYNC_POINT_CALLBACK("CompactionPicker::PickFilesMarkedForCompaction",
&random_file_index);
if (continuation(vstorage->FilesMarkedForCompaction()[random_file_index])) {
// found the compaction!
return;
}
for (auto& level_file : vstorage->FilesMarkedForCompaction()) {
if (continuation(level_file)) {
// found the compaction!
return;
}
}
start_level_inputs->files.clear();
}
bool CompactionPicker::GetOverlappingL0Files(
VersionStorageInfo* vstorage, CompactionInputFiles* start_level_inputs,
int output_level, int* parent_index) {
// Two level 0 compaction won't run at the same time, so don't need to worry
// about files on level 0 being compacted.
assert(level0_compactions_in_progress()->empty());
InternalKey smallest, largest;
GetRange(*start_level_inputs, &smallest, &largest);
// Note that the next call will discard the file we placed in
// c->inputs_[0] earlier and replace it with an overlapping set
// which will include the picked file.
start_level_inputs->files.clear();
vstorage->GetOverlappingInputs(0, &smallest, &largest,
&(start_level_inputs->files));
// If we include more L0 files in the same compaction run it can
// cause the 'smallest' and 'largest' key to get extended to a
// larger range. So, re-invoke GetRange to get the new key range
GetRange(*start_level_inputs, &smallest, &largest);
if (IsRangeInCompaction(vstorage, &smallest, &largest, output_level,
parent_index)) {
return false;
}
assert(!start_level_inputs->files.empty());
return true;
}
} // namespace ROCKSDB_NAMESPACE