Summary:
Currently we rely on FilterBitsBuilder to skip the duplicate keys. It does that by comparing that hash of the key to the hash of the last added entry. This logic breaks however when we have whole_key_filtering mixed with prefix blooms as their addition to FilterBitsBuilder will be interleaved. The patch fixes that by comparing the last whole key and last prefix with the whole key and prefix of the new key respectively and skip the call to FilterBitsBuilder if it is a duplicate.
Closes https://github.com/facebook/rocksdb/pull/3764
Differential Revision: D7744413
Pulled By: maysamyabandeh
fbshipit-source-id: 15df73bbbafdfd754d4e1f42ea07f47b03bc5eb8
Summary:
The WriteBatch::Iterate will try with a larger sequence number if the memtable reports a duplicate. This status is specified with TryAgain status. So far the assumption was that the last entry in the batch will never return TryAgain, which is correct when WAL is created via WritePrepared since it always appends a batch separator if a natural one does not exist. However when reading a WAL generated by WriteCommitted this batch separator might not exist. Although WritePrepared is not supposed to be able to read the WAL generated by WriteCommitted we should avoid confusing scenarios in which the behavior becomes unpredictable. The path fixes that by allowing TryAgain even for the last entry of the write batch.
Closes https://github.com/facebook/rocksdb/pull/3747
Differential Revision: D7708391
Pulled By: maysamyabandeh
fbshipit-source-id: bfaddaa9b14a4cdaff6977f6f63c789a6ab1ee0d
Summary:
This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic.
Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction)
This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2.
Closes https://github.com/facebook/rocksdb/pull/3765
Differential Revision: D7747618
Pulled By: siying
fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
Summary:
Ttl-triggered and snapshot-release-triggered compactions should not be considered as manual compactions. This is a bug.
Closes https://github.com/facebook/rocksdb/pull/3678
Differential Revision: D7498151
Pulled By: sagar0
fbshipit-source-id: a2d5bed05268a4dc93d54ea97a9ae44b366df15d
Summary:
Level Compaction with TTL.
As of today, a file could exist in the LSM tree without going through the compaction process for a really long time if there are no updates to the data in the file's key range. For example, in certain use cases, the keys are not actually "deleted"; instead they are just set to empty values. There might not be any more writes to this "deleted" key range, and if so, such data could remain in the LSM for a really long time resulting in wasted space.
Introducing a TTL could solve this problem. Files (and, in turn, data) older than TTL will be scheduled for compaction when there is no other background work. This will make the data go through the regular compaction process and get rid of old unwanted data.
This also has the (good) side-effect of all the data in the non-bottommost level being newer than ttl, and all data in the bottommost level older than ttl. It could lead to more writes while reducing space.
This functionality can be controlled by the newly introduced column family option -- ttl.
TODO for later:
- Make ttl mutable
- Extend TTL to Universal compaction as well? (TTL is already supported in FIFO)
- Maybe deprecate CompactionOptionsFIFO.ttl in favor of this new ttl option.
Closes https://github.com/facebook/rocksdb/pull/3591
Differential Revision: D7275442
Pulled By: sagar0
fbshipit-source-id: dcba484717341200d419b0953dafcdf9eb2f0267
Summary:
This is a hack as temporary fix of MyRocks with rollbacking the merge operands. The way MyRocks uses merge operands is without protection of locks, which violates the assumption behind the rollback algorithm. They are ok with not being rolled back as it would just create a gap in the autoincrement column. The hack add an option to disable the rollback of merge operands by default and only enables it to let the unit test pass.
Closes https://github.com/facebook/rocksdb/pull/3711
Differential Revision: D7597177
Pulled By: maysamyabandeh
fbshipit-source-id: 544be0f666c7e7abb7f651ec8b23124e05056728
Summary:
We introduced smallest_prep optimization in this commit b225de7e10, which enables storing the smallest uncommitted sequence number along with the snapshot. This enables the readers that read from the snapshot to skip further checks and safely assumed the data is committed if its sequence number is less than smallest uncommitted when the snapshot was taken. The problem was that smallest uncommitted and the snapshot must be taken atomically, and the lack of atomicity had led to readers using a smallest uncommitted after the snapshot was taken and hence mistakenly skipping some data.
This patch fixes the problem by i) separating the process of removing of prepare entries from the AddCommitted function, ii) removing the prepare entires AFTER the committed sequence number is published, iii) getting smallest uncommitted (from the prepare list) BEFORE taking a snapshot. This guarantees that the smallest uncommitted that is accompanied with a snapshot is less than or equal of such number if it was obtained atomically.
Tested by running MySQLStyleTransactionTest/MySQLStyleTransactionTest.TransactionStressTest that was failing sporadically.
Closes https://github.com/facebook/rocksdb/pull/3703
Differential Revision: D7581934
Pulled By: maysamyabandeh
fbshipit-source-id: dc9d6f4fb477eba75d4d5927326905b548a96a32
Summary:
The existing unit test did not set the level so the check for pinned partitioned filter/index being properly released from the block cache was not properly exercised as they only take effect in level 0. As a result a memory leak in pinned partitioned filters was hidden. The patch fix the test as well as the bug.
Closes https://github.com/facebook/rocksdb/pull/3692
Differential Revision: D7559763
Pulled By: maysamyabandeh
fbshipit-source-id: 55eff274945838af983c764a7d71e8daff092e4a
Summary:
Adding some stats that would be helpful to monitor if the DB has gone to unlikely stats that would hurt the performance. These are mostly when we end up needing to acquire a mutex.
Closes https://github.com/facebook/rocksdb/pull/3683
Differential Revision: D7529393
Pulled By: maysamyabandeh
fbshipit-source-id: f7d36279a8f39bd84d8ddbf64b5c97f670c5d6d9
Summary:
Currently dump_wal cannot print the prepared records from the WAL that is generated by WRITE_PREPARED write policy since the default reaction of the handler is to return NotSupported if markers of WRITE_PREPARED are encountered. This patch enables the admin to pass --write_committed=false option, which will be accordingly passed to the handler. Note that DBFileDumperCommand and DBDumperCommand are still not updated by this patch but firstly they are not urgent and secondly we need to revise this approach later when we also add WRITE_UNPREPARED markers so I leave it for future work.
Tested by running it on a WAL generated by WRITE_PREPARED:
$ ./ldb dump_wal --walfile=/dev/shm/dbbench/000003.log | grep BEGIN_PREARE | head -1
1,2,70,0,BEGIN_PREARE
$ ./ldb dump_wal --walfile=/dev/shm/dbbench/000003.log --write_committed=false | grep BEGIN_PREARE | head -1
1,2,70,0,BEGIN_PREARE PUT(0) : 0x30303031313330313938 PUT(0) : 0x30303032353732313935 END_PREPARE(0x74786E31313535383434323738303738363938313335312D30)
Closes https://github.com/facebook/rocksdb/pull/3682
Differential Revision: D7522090
Pulled By: maysamyabandeh
fbshipit-source-id: a0332207261c61e18b2f9dfbe9feecd9a1339aca
Summary:
Adds two stats to allow us measuring the false positive rate of full filters:
- The total count of positives: rocksdb.bloom.filter.full.positive
- The total count of true positives: rocksdb.bloom.filter.full.true.positive
Not the term "full" in the stat name to indicate that they are meaningful in full filters. block-based filters are to be deprecated soon and supporting it is not worth the the additional cost of if-then-else branches.
Closes#3680
Tested by:
$ ./db_bench -benchmarks=fillrandom -db /dev/shm/rocksdb-tmpdb --num=1000000 -bloom_bits=10
$ ./db_bench -benchmarks="readwhilewriting" -db /dev/shm/rocksdb-tmpdb --statistics -bloom_bits=10 --duration=60 --num=2000000 --use_existing_db 2>&1 > /tmp/full.log
$ grep filter.full /tmp/full.log
rocksdb.bloom.filter.full.positive COUNT : 3628593
rocksdb.bloom.filter.full.true.positive COUNT : 3536026
which gives the false positive rate of 2.5%
Closes https://github.com/facebook/rocksdb/pull/3681
Differential Revision: D7517570
Pulled By: maysamyabandeh
fbshipit-source-id: 630ab1a473afdce404916d297035b6318de4c052
Summary:
The is an optimization to reduce lookup in the CommitCache when querying IsInSnapshot. The optimization takes the smallest uncommitted data at the time that the snapshot was taken and if the sequence number of the read data is lower than that number it assumes the data as committed.
To implement this optimization two changes are required: i) The AddPrepared function must be called sequentially to avoid out of order insertion in the PrepareHeap (otherwise the top of the heap does not indicate the smallest prepare in future too), ii) non-2PC transactions also call AddPrepared if they do not commit in one step.
Closes https://github.com/facebook/rocksdb/pull/3649
Differential Revision: D7388630
Pulled By: maysamyabandeh
fbshipit-source-id: b79506238c17467d590763582960d4d90181c600
Summary:
This patch record the deleted WAL numbers in the manifest to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic.
Closes https://github.com/facebook/rocksdb/pull/3488
Differential Revision: D6967893
Pulled By: maysamyabandeh
fbshipit-source-id: 13119feb155a08ab6d4909f437c7a750480dc8a1
Summary:
When using two_write_queue, the published seq and the last allocated sequence could be ahead of the LastSequence, even if both write queues are stopped as in WriteRecoverableState. The patch fixes a bug in WriteRecoverableState in which LastSequence was used as a reference but the result was applied to last fetched sequence and last published seq.
Closes https://github.com/facebook/rocksdb/pull/3665
Differential Revision: D7446099
Pulled By: maysamyabandeh
fbshipit-source-id: 1449bed9aed8e9db6af85946efd347cb8efd3c0b
Summary:
Currently if the CommitTimeWriteBatch is set to be used only as a state that is required only for recovery , the user cannot see that in DB until it is restarted. This while the state is already inserted into the DB after the memtable flush. It would be useful for debugging if make this state visible to the user after the flush by committing it. The patch does it by a invoking a callback that does the commit on the recoverable state.
Closes https://github.com/facebook/rocksdb/pull/3661
Differential Revision: D7424577
Pulled By: maysamyabandeh
fbshipit-source-id: 137f9408662f0853938b33fa440f27f04c1bbf5c
Summary:
Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst.
Also add a unit test to verify that this PR fixes the issue.
The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows
```
timeline user_thread background_compaction thread
t1 | FindObsoleteFiles(full_scan=false)
t2 | FindObsoleteFiles(full_scan=true)
t3 | PurgeObsoleteFiles
t4 | PurgeObsoleteFiles
V
```
When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`.
To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files.
ajkr could you take a look and comment? Thanks!
Closes https://github.com/facebook/rocksdb/pull/3638
Differential Revision: D7384372
Pulled By: riversand963
fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
Summary:
Updated comments around AllowSingleOperand.
Reason: A couple of users were confused and encountered issues due to no overriding PartialMerge with AllowSingleOperand=true.
I'll also look into modifying the default merge operator implementation so that overriding PartialMerge is not mandatory when AllowSingleOp=true.
Closes https://github.com/facebook/rocksdb/pull/3659
Differential Revision: D7422691
Pulled By: sagar0
fbshipit-source-id: 3d075a6ced0120f5d65cb7ae5412936f1862f342
Summary:
Our valgrind continuous test found an interesting leak which got introduced in #3614. We were adding the prefix key before saving the previous prefix start offset, due to which previous prefix offset is always incorrect. Fixed it by saving the the previous sate before adding the key.
Closes https://github.com/facebook/rocksdb/pull/3660
Differential Revision: D7418698
Pulled By: sagar0
fbshipit-source-id: 9933685f943cf2547ed5c553f490035a2fa785cf
Summary:
Provide a block_align option in BlockBasedTableOptions to allow
alignment of SST file data blocks. This will avoid higher
IOPS/throughput load due to < 4KB data blocks spanning 2 4KB pages.
When this option is set to true, the block alignment is set to lower of
block size and 4KB.
Closes https://github.com/facebook/rocksdb/pull/3502
Differential Revision: D7400897
Pulled By: anand1976
fbshipit-source-id: 04cc3bd144e88e3431a4f97604e63ad7a0f06d44
Summary:
Current commit cache size is 2^21. This was due to a type. With 2^23 commit entries we can have transactions as long as 64s without incurring the cost of having them evicted from the commit cache before their commit. Here is the math:
2^23 / 2 (one out of two seq numbers are for commit) / 2^16 TPS = 2^6 = 64s
Closes https://github.com/facebook/rocksdb/pull/3657
Differential Revision: D7411211
Pulled By: maysamyabandeh
fbshipit-source-id: e7cacf40579f3acf940643d8a1cfe5dd201caa35
Summary:
Currently log_writer->AddRecord in WriteImpl is protected from concurrent calls via FlushWAL only if two_write_queues_ option is set. The patch fixes the problem by i) skip log_writer->AddRecord in FlushWAL if manual_wal_flush is not set, ii) protects log_writer->AddRecord in WriteImpl via log_write_mutex_ if manual_wal_flush_ is set but two_write_queues_ is not.
Fixes#3599
Closes https://github.com/facebook/rocksdb/pull/3656
Differential Revision: D7405608
Pulled By: maysamyabandeh
fbshipit-source-id: d6cc265051c77ae49c7c6df4f427350baaf46934
Summary:
I found that each instance of MySQLStyleTransactionTest.TransactionStressTest/x is taking more than 10 hours to complete on our continuous testing environment, causing the whole valgrind run to timeout after a day. So excluding these tests.
Closes https://github.com/facebook/rocksdb/pull/3652
Differential Revision: D7400332
Pulled By: sagar0
fbshipit-source-id: 987810574506d01487adf7c2de84d4817ec3d22d
Summary:
Currently AddPrepared is performed only on the first sub-batch if there are duplicate keys in the write batch. This could cause a problem if the transaction takes too long to commit and the seq number of the first sub-patch moved to old_prepared_ but not the seq of the later ones. The patch fixes this by calling AddPrepared for all sub-patches.
Closes https://github.com/facebook/rocksdb/pull/3651
Differential Revision: D7388635
Pulled By: maysamyabandeh
fbshipit-source-id: 0ccd80c150d9bc42fe955e49ddb9d7ca353067b4
Summary:
RangeDelAggregator will remember the files whose range tombstones have been added,
so the caller can check whether the file has been added before call AddTombstones.
Closes https://github.com/facebook/rocksdb/pull/3635
Differential Revision: D7354604
Pulled By: ajkr
fbshipit-source-id: 9b9f7ec130556028df417e650711554b46d8d107
Summary:
We have not been updating our HISTORY.md change log with the RocksJava changes. Going forward, lets add Java changes also to HISTORY.md.
There is an old java/HISTORY-JAVA.md, but it hasn't been updated in years. It is much easier to remember to update the change log in a single file, HISTORY.md.
I added information about shared block cache here, which was introduced in #3623.
Closes https://github.com/facebook/rocksdb/pull/3647
Differential Revision: D7384448
Pulled By: sagar0
fbshipit-source-id: 9b6e569f44e6df5cb7ba06413d9975df0b517d20
Summary:
Summary
========
`InlineSkipList<>::Insert` takes the `key` parameter as a C-string. Then, it performs multiple comparisons with it requiring the `GetLengthPrefixedSlice()` to be spawn in `MemTable::KeyComparator::operator()(const char* prefix_len_key1, const char* prefix_len_key2)` on the same data over and over. The patch tries to optimize that.
Rough performance comparison
=====
Big keys, no compression.
```
$ ./db_bench --writes 20000000 --benchmarks="fillrandom" --compression_type none -key_size 256
(...)
fillrandom : 4.222 micros/op 236836 ops/sec; 80.4 MB/s
```
```
$ ./db_bench --writes 20000000 --benchmarks="fillrandom" --compression_type none -key_size 256
(...)
fillrandom : 4.064 micros/op 246059 ops/sec; 83.5 MB/s
```
TODO
======
In ~~a separated~~ this PR:
- [x] Go outside the write path. Maybe even eradicate the C-string-taking variant of `KeyIsAfterNode` entirely.
- [x] Try to cache the transformations applied by `KeyComparator` & friends in situations where we havy many comparisons with the same key.
Closes https://github.com/facebook/rocksdb/pull/3516
Differential Revision: D7059300
Pulled By: ajkr
fbshipit-source-id: 6f027dbb619a488129f79f79b5f7dbe566fb2dbb
Summary:
Fsync after writing global sequence number to the ingestion file in ExternalSstFileIngestionJob. Otherwise the file metadata could be incorrect.
Closes https://github.com/facebook/rocksdb/pull/3644
Differential Revision: D7373813
Pulled By: sagar0
fbshipit-source-id: 4da2c9e71a8beb5c08b4ac955f288ee1576358b8
Summary:
It was misnamed. It actually updates `bg_error_` if `PreprocessWrite()` or `WriteToWAL()` fail, not related to the user callback.
Closes https://github.com/facebook/rocksdb/pull/3485
Differential Revision: D6955787
Pulled By: ajkr
fbshipit-source-id: bd7afc3fdb7a52830c021cbfc25fcbc3ab7d5e10
Summary:
Add `bytes_max_delete_chunk` in SstFileManager so that we can drop a large file in multiple batches.
Closes https://github.com/facebook/rocksdb/pull/3640
Differential Revision: D7358679
Pulled By: siying
fbshipit-source-id: ef17f0da2f5723dbece2669485a9b91b3edc0bb7
Summary:
This commit fixes a race condition on calling SetLastPublishedSequence. The function must be called only from the 2nd write queue when two_write_queues is enabled. However there was a bug that would also call it from the main write queue if CommitTimeWriteBatch is provided to the commit request and yet use_only_the_last_commit_time_batch_for_recovery optimization is not enabled. To fix that we penalize the commit request in such cases by doing an additional write solely to publish the seq number from the 2nd queue.
Closes https://github.com/facebook/rocksdb/pull/3641
Differential Revision: D7361508
Pulled By: maysamyabandeh
fbshipit-source-id: bf8f7a27e5cccf5425dccbce25eb0032e8e5a4d7
Summary:
The 10MB buffer in BackupEngineImpl::BackupMeta::StoreToFile can be corrupted with a large number of files. Added a check to determine current buffer length and append data to file if buffer becomes full.
Resolves https://github.com/facebook/rocksdb/issues/3228
Closes https://github.com/facebook/rocksdb/pull/3636
Differential Revision: D7354160
Pulled By: ajkr
fbshipit-source-id: eec12d38095a0d17551a4aaee52b99d30a555722
Summary:
This pull request exposes the interface of PerfContext as C API
Closes https://github.com/facebook/rocksdb/pull/3607
Differential Revision: D7294225
Pulled By: ajkr
fbshipit-source-id: eddcfbc13538f379950b2c8b299486695ffb5e2c
Summary:
Changes to support sharing block cache using the Java API.
Previously DB instances could share the block cache only when the same Options instance is passed to all the DB instances. But now, with this change, it is possible to explicitly create a cache and pass it to multiple options instances, to share the block cache.
Implementing this for [Rocksandra](https://github.com/instagram/cassandra/tree/rocks_3.0), but this feature has been requested by many java api users over the years.
Closes https://github.com/facebook/rocksdb/pull/3623
Differential Revision: D7305794
Pulled By: sagar0
fbshipit-source-id: 03e4e8ed7aeee6f88bada4a8365d4279ede2ad71
Summary:
* Fix BlobDBImpl::GCFileAndUpdateLSM doesn't close the new file, and the new file will not be able to be garbage collected later.
* Fix BlobDBImpl::GCFileAndUpdateLSM doesn't copy over metadata from old file to new file.
Closes https://github.com/facebook/rocksdb/pull/3639
Differential Revision: D7355092
Pulled By: yiwu-arbug
fbshipit-source-id: 4fa3594ac5ce376bed1af04a545c532cfc0088c4
Summary:
When destorying column family handle after the column family has been deleted, the handle may hold share pointers of some objects in ColumnFamilyOptions, but in the destructor, the destructing order may cause some of the objects to be destoryed before being used by the following steps. Fix it by making a copy of the option object and destory it as the last step.
Closes https://github.com/facebook/rocksdb/pull/3610
Differential Revision: D7281025
Pulled By: siying
fbshipit-source-id: ac18f3b2841788cba4ccfa1abd8d59158c1113bc
