Summary:
In the write path, there is an optimization: when a new WAL is created during SwitchMemtable, we update the internal log number of the empty column families to the new WAL. `FindObsoleteFiles` marks a WAL as obsolete if the WAL's log number is less than `VersionSet::MinLogNumberWithUnflushedData`. After updating the empty column families' internal log number, `VersionSet::MinLogNumberWithUnflushedData` might change, so some WALs might become obsolete to be purged from disk.
For example, consider there are 3 column families: 0, 1, 2:
1. initially, all the column families' log number is 1;
2. write some data to cf0, and flush cf0, but the flush is pending;
3. now a new WAL 2 is created;
4. write data to cf1 and WAL 2, now cf0's log number is 1, cf1's log number is 2, cf2's log number is 2 (because cf1 and cf2 are empty, so their log numbers will be set to the highest log number);
5. now cf0's flush hasn't finished, flush cf1, a new WAL 3 is created, and cf1's flush finishes, now cf0's log number is 1, cf1's log number is 3, cf2's log number is 3, since WAL 1 still contains data for the unflushed cf0, no WAL can be deleted from disk;
6. now cf0's flush finishes, cf0's log number is 2 (because when cf0 was switching memtable, WAL 3 does not exist yet), cf1's log number is 3, cf2's log number is 3, so WAL 1 can be purged from disk now, but WAL 2 still cannot because `MinLogNumberToKeep()` is 2;
7. write data to cf2 and WAL 3, because cf0 is empty, its log number is updated to 3, so now cf0's log number is 3, cf1's log number is 3, cf2's log number is 3;
8. now if the background threads want to purge obsolete files from disk, WAL 2 can be purged because `MinLogNumberToKeep()` is 3. But there are only two flush results written to MANIFEST: the first is for flushing cf1, and the `MinLogNumberToKeep` is 1, the second is for flushing cf0, and the `MinLogNumberToKeep` is 2. So without this PR, if the DB crashes at this point and try to recover, `WalSet` will still expect WAL 2 to exist.
When WAL tracking is enabled, we assume WALs will only become obsolete after a flush result is written to MANIFEST in `MemtableList::TryInstallMemtableFlushResults` (or its atomic flush counterpart). The above situation breaks this assumption.
This PR tracks WAL obsoletion if necessary before updating the empty column families' log numbers.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7781
Test Plan:
watch existing tests and stress tests to pass.
`make -j48 blackbox_crash_test` on devserver
Reviewed By: ltamasi
Differential Revision: D25631695
Pulled By: cheng-chang
fbshipit-source-id: ca7fff967bdb42204b84226063d909893bc0a4ec
Summary:
When ConcurrentTaskLimiter is enabled and there are too many outstanding compactions, BackgroundCompaction returns Status::Busy(), which shouldn't be treat as compaction failure.
This caused performance issue when outstanding compactions reached the limit.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7739
Reviewed By: cheng-chang
Differential Revision: D25508319
Pulled By: ltamasi
fbshipit-source-id: 3b181b16ada0ca3393cfa3a7412985764e79c719
Summary:
sst file number in corruption error would be very useful for debugging
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7767
Reviewed By: zhichao-cao
Differential Revision: D25485872
Pulled By: jay-zhuang
fbshipit-source-id: 67315b582cedeefbce6676015303ebe5bf6526a3
Summary:
The patch adds initial support for reading blobs to the batched `MultiGet` API.
The current implementation simply retrieves the blob values as the blob indexes
are encountered; that is, reads from blob files are currently not batched. (This
will be optimized in a separate phase.) In addition, the patch removes some dead
code related to BlobDB from the batched `MultiGet` implementation, namely the
`is_blob` / `is_blob_index` flags that are passed around in `DBImpl` and `MemTable` /
`MemTableListVersion`. These were never hooked up to anything and wouldn't
work anyways, since a single flag is not sufficient to communicate the "blobness"
of multiple key-values.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7766
Test Plan: `make check`
Reviewed By: jay-zhuang
Differential Revision: D25479290
Pulled By: ltamasi
fbshipit-source-id: 7aba2d290e31876ee592bcf1adfd1018713a8000
Summary:
Uncommon bug seen by ASAN with
ColumnFamilyTest.LiveIteratorWithDroppedColumnFamily, if the last two
references to a ColumnFamilyData are both SuperVersions (during
InstallSuperVersion). The fix is to use UnrefAndTryDelete even in
SuperVersion::Cleanup but with a parameter to avoid re-entering Cleanup
on the same SuperVersion being cleaned up.
ColumnFamilyData::Unref is considered unsafe so removed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7749
Test Plan: ./column_family_test --gtest_filter=*LiveIter* --gtest_repeat=100
Reviewed By: jay-zhuang
Differential Revision: D25354304
Pulled By: pdillinger
fbshipit-source-id: e78f3a3f67c40013b8432f31d0da8bec55c5321c
Summary:
min_wal_number_to_keep should not be decreasing, if it does not increase, then there is no need to log the WAL obsoletions in MANIFEST since a previous one has been logged.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7765
Test Plan: watch existing tests and stress tests to pass
Reviewed By: pdillinger
Differential Revision: D25462542
Pulled By: cheng-chang
fbshipit-source-id: 0085fcb6edf5cf2b0fc32f9932a7566f508768ff
Summary:
When two phase commit is enabled, `VersionSet::min_log_number_to_keep_2pc` is set during flush.
But when a new MANIFEST is created, the `min_log_number_to_keep_2pc` is not carried over to the new MANIFEST. So if a new MANIFEST is created and then DB is reopened, the `min_log_number_to_keep_2pc` will be lost. This may cause DB recovery errors.
The bug is reproduced in a new unit test in `version_set_test.cc`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7747
Test Plan: The new unit test in `version_set_test.cc` should pass.
Reviewed By: jay-zhuang
Differential Revision: D25350661
Pulled By: cheng-chang
fbshipit-source-id: eee890d5b19f15769069670692e270ae31044ece
Summary:
Ensure that when direct IO is enabled and a compressed block cache is
configured, MultiGet inserts compressed data blocks into the compressed
block cache.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7756
Test Plan: Add unit test to db_basic_test
Reviewed By: cheng-chang
Differential Revision: D25416240
Pulled By: anand1976
fbshipit-source-id: 75d57526370c9c0a45ff72651f3278dbd8a9086f
Summary:
If WAL tracking was enabled, then disabled during reopen, the previously tracked WALs should be removed from MANIFEST.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7757
Test Plan: a new unit test `DBBasicTest.DisableTrackWal` is added.
Reviewed By: jay-zhuang
Differential Revision: D25410508
Pulled By: cheng-chang
fbshipit-source-id: 9d8d9e665066135930a7c1035bb8c2f68bded6a0
Summary:
Currently, when a WAL becomes obsolete after flushing, if VersionSet::WalSet does not contain the WAL, we do not track the WAL obsoletion event in MANIFEST.
But consider this case:
* WAL 10 is synced, a VersionEdit is LogAndApplied to MANIFEST to log this WAL addition event, but the VersionEdit is not applied to WalSet yet since its corresponding ManifestWriter is still pending in the write queue;
* Since the above ManifestWriter is blocking, the LogAndApply will block on a conditional variable and release the db mutex, so another LogAndApply can proceed to enqueue other VersionEdits concurrently;
* Now flush happens, and WAL 10 becomes obsolete, although WalSet does not contain WAL 10 yet, we should call LogAndApply to enqueue a VersionEdit to indicate the obsoletion of WAL 10;
* otherwise, when the queued edit indicating WAL 10 addition is logged to MANIFEST, and DB crashes and reopens, the WAL 10 might have been removed from disk, but it still exists in MANIFEST.
This PR changes the behavior to: always `LogAndApply` any WAL addition or obsoletion event, without considering the order issues caused by concurrency, but when applying the edits to `WalSet`, do not add the WALs if they are already obsolete. In this approach, the logical events of WAL addition and obsoletion are always tracked in MANIFEST, so we can inspect the MANIFEST and know all the previous WAL events, but we choose to ignore certain events due to the concurrency issues such as the case above, or the case in https://github.com/facebook/rocksdb/pull/7725.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7759
Test Plan: make check
Reviewed By: pdillinger
Differential Revision: D25423089
Pulled By: cheng-chang
fbshipit-source-id: 9cb9a7fbc1875bf954f2a42f9b6cfd6d49a7b21c
Summary:
Consider the case:
1. All column families are flushed, so all WALs become obsolete, but no WAL is removed from disk yet because the removal is asynchronous, a VersionEdit is written to MANIFEST indicating that WALs before a certain WAL number are obsolete, let's say this number is 3;
2. `SyncWAL` is called, so all the on-disk WALs are synced, and if track_and_verify_wal_in_manifest=true, the WALs will be tracked in MANIFEST, let's say the WAL numbers are 1 and 2;
3. DB crashes;
4. During DB recovery, when replaying MANIFEST, we first see that WAL with number < 3 are obsolete, then we see that WAL 1 and 2 are synced, so according to current implementation of `WalSet`, the `WalSet` will be recovered to include WAL 1 and 2;
5. WAL 1 and 2 are asynchronously deleted from disk, then the WAL verification algorithm fails with `Corruption: missing WAL`.
The above case is reproduced in a new unit test `DBBasicTestTrackWal::DoNotTrackObsoleteWal`.
The fix is to maintain the upper bound of the obsolete WAL numbers, any WAL with number less than the maintained number is considered to be obsolete, so shouldn't be tracked even if they are later synced. The number is maintained in `WalSet`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7725
Test Plan:
1. a new unit test `DBBasicTestTrackWal::DoNotTrackObsoleteWal` is added.
2. run `make crash_test` on devserver.
Reviewed By: riversand963
Differential Revision: D25238914
Pulled By: cheng-chang
fbshipit-source-id: f5dccd57c3d89f19565ec5731f2d42f06d272b72
Summary:
This PR removes a nested loop inside ProcessManifestWrites. The new
implementation has the same behavior as the old code with simpler logic
and lower complexity.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7751
Test Plan:
make check
Run make crash_test on devserver and succeeds 3 times.
Reviewed By: ltamasi
Differential Revision: D25363526
Pulled By: riversand963
fbshipit-source-id: 27e681949dacd7501a752e5e517b9e85b54ccb2e
Summary:
This change eliminates the need for a lot of the PermitUncheckedError calls on return from ErrorHandler methods. The calls are no longer needed as the status is returned as a reference rather than a copy. Additionally, this means that the originating status (recovery_error_, bg_error_) is not cleared implicitly as a result of calling one of these methods.
For this class, I do not know if the proper behavior should be to call PermitUncheckedError in the destructor or if the checked state should be cleared when the status is cleared. I did tests both ways. Without the code in the destructor, the status will need to be cleared in at least some of the places where it is set to OK. When running tests, I found no instances where this class was destructed with a non-OK, non-checked Status.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7539
Reviewed By: anand1976
Differential Revision: D25340565
Pulled By: pdillinger
fbshipit-source-id: 1730c035c81a475875ea745226112030ec25136c
Summary:
`googletest` uses exceptions to communicate assertion failures when
`GTEST_THROW_ON_FAILURE` is set, which does not go well with
`std::thread`s, since an exception escaping the top-level function of an
`std::thread` object or an `std::thread` getting destroyed without
having been `join`ed or `detach`ed first results in a call to
`std::terminate`. The patch fixes this by moving the `Status` assertions
of background operations in `ExternalSstFileTest.PickedLevelBug` to the
main thread.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7754
Test Plan: `make check`
Reviewed By: riversand963
Differential Revision: D25383808
Pulled By: ltamasi
fbshipit-source-id: 32fb2721e5169ec898d218900bc0d83eead45d03
Summary:
Handle misuse of snprintf return value to avoid Out of bound
read/write.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7686
Test Plan: make check -j64
Reviewed By: riversand963
Differential Revision: D25030831
Pulled By: akankshamahajan15
fbshipit-source-id: 1a1d181c067c78b94d720323ae00b79566b57cfa
Summary:
Added a fix for the failure of
DBTest2.PartitionedIndexUserToInternalKey on ppc64le in travis
Closes https://github.com/facebook/rocksdb/issues/7746
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7752
Test Plan:
Ran travis job multiple times and it passed. Will keep
watching the travis job after this patch.
Reviewed By: pdillinger
Differential Revision: D25373130
Pulled By: akankshamahajan15
fbshipit-source-id: fa0e3f85f75b687415044a506e42cc38ead87975
Summary:
Following https://github.com/facebook/rocksdb/issues/7655 and https://github.com/facebook/rocksdb/issues/7657, this PR adds `full_history_ts_low_` to `ColumnFamilyData`.
`ColumnFamilyData::full_history_ts_low_` will be used to create `FlushJob` and `CompactionJob`.
`ColumnFamilyData::full_history_ts_low` is persisted to the MANIFEST file. An application can only
increase its value. Consider the following case:
>
> The database has a key at ts=950. `full_history_ts_low` is first set to 1000, and then a GC is triggered
> and cleans up all data older than 1000. If the application sets `full_history_ts_low` to 900 afterwards,
> and tries to read at ts=960, the key at 950 is not seen. From the perspective of the read, the result
> is hard to reason. For simplicity, we just do now allow decreasing full_history_ts_low for now.
>
During recovery, the value of `full_history_ts_low` is restored for each column family if applicable. Note that
version edits in the MANIFEST file for the same column family may have `full_history_ts_low` unsorted due
to the potential interleaving of `LogAndApply` calls. Only the max will be used to restore the state of the
column family.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7740
Test Plan: make check
Reviewed By: ltamasi
Differential Revision: D25296217
Pulled By: riversand963
fbshipit-source-id: 24acda1df8262cd7cfdc6ce7b0ec56438abe242a
Summary:
The patch adds iterator support to the integrated BlobDB implementation.
Whenever a blob reference is encountered during iteration, the corresponding
blob is retrieved by calling `Version::GetBlob`, assuming the `expose_blob_index`
(formerly `allow_blob`) flag is *not* set. (Note: the flag is set by the old stacked
BlobDB implementation, which has its own blob file handling/blob retrieval logic.)
In addition, `DBIter` now uniformly returns `Status::NotSupported` with the error
message `"BlobDB does not support merge operator."` when encountering a
blob reference while performing a merge (instead of potentially returning a
message that implies the database should be opened using the stacked BlobDB's
`Open`.)
TODO: We can implement support for lazily retrieving the blob value (or in other
words, bypassing the retrieval of blob values based on key) by extending the `Iterator`
API with a new `PrepareValue` method (similarly to `InternalIterator`, which already
supports lazy values).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7731
Test Plan: `make check`
Reviewed By: riversand963
Differential Revision: D25256293
Pulled By: ltamasi
fbshipit-source-id: c39cd782011495a526cdff99c16f5fca400c4811
Summary:
In current code base, in FlushMemtable, when `(Flush_reason == FlushReason::kErrorRecoveryRetryFlush && (!cfd->mem()->IsEmpty() || !cached_recoverable_state_empty_.load()))`, we assert that cfd->imm()->NumNotFlushed() > 0. However, there are some corner cases that can fail this assert: 1) if there are multiple CFs, some CF has immutable memtable, some CFs don't. In ResumeImpl, all CFs will call FlushMemtable, which will hit the assert. 2) Regular flush is scheduled and running, the resume thread is waiting. New KVs are inserted and SchedulePendingFlush is called. Regular flush will continue call MaybeScheduleFlushAndCompaction until all the immutable memtables are flushed. When regular flush ends and auto resume thread starts to schedule new flushes, cfd->imm()->NumNotFlushed() can be 0.
Remove the assert and added the comments.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7744
Test Plan: make check and pass the stress test
Reviewed By: riversand963
Differential Revision: D25340573
Pulled By: zhichao-cao
fbshipit-source-id: eac357bdace660247c197f01a9ff6857e3c97672
Summary:
In error_handler auto recovery case, if recovery_in_prog_ is false, the recover is finished or failed. In this case, the auto recovery thread should finish its execution so recovery_thread_ should be null. However, in some cases, it is not null, the caller should not directly returned. Instead, it should wait for a while and create a new thread to execute the new recovery.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7700
Test Plan: make check, error_handler_fs_test
Reviewed By: anand1976
Differential Revision: D25098233
Pulled By: zhichao-cao
fbshipit-source-id: 5a1cba234ca18f6dd5d1be88e02d66e1d5ce931b
Summary:
When 2 phase commit is enabled, if there are prepared data in a WAL, the WAL should be kept, the minimum log number for such a WAL is written to MANIFEST during flush. In atomic flush, such information is not written to MANIFEST.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7570
Test Plan: Added a new unit test `DBAtomicFlushTest.ManualFlushUnder2PC`, this test fails in atomic flush without this PR, after this PR, it succeeds.
Reviewed By: riversand963
Differential Revision: D24394222
Pulled By: cheng-chang
fbshipit-source-id: 60ce74b21b704804943be40c8de01b41269cf116
Summary:
In the current code base, all the manifest writes with IO error will be set with reason: BackgroundErrorReason::kManifestWrite, which will be mapped to the kHardError if the IO Error is retryable. However, if the system does not use the WAL, all the retryable IO error should be mapped to kSoftError. Create this PR to handle is special case by adding kManifestWriteNoWAL to BackgroundErrorReason.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7693
Test Plan: make check, add new testing cases to error_handler_fs_test
Reviewed By: anand1976
Differential Revision: D25066204
Pulled By: zhichao-cao
fbshipit-source-id: d59553896c2eac3fb37c05238544d2b265379462
Summary:
Add timestamp to the `CompactRange()` and `GetApproximateSizes` range keys if needed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7684
Test Plan: make check
Reviewed By: riversand963
Differential Revision: D25015421
Pulled By: jay-zhuang
fbshipit-source-id: 51ca0756087eb053a3b11801e5c7ce1c6e2d38a9
Summary:
https://github.com/facebook/rocksdb/issues/7340 reports and reproduces an assertion failure caused by a combination of the following:
- atomic flush is disabled.
- a column family can appear multiple times in the flush queue at the same time. This behavior was introduced in release 5.17.
Consequently, it is possible that two flushes race with each other. One bg flush thread flushes all memtables. The other thread calls `FlushMemTableToOutputFile()` afterwards, and hits the assertion error below.
```
assert(cfd->imm()->NumNotFlushed() != 0);
assert(cfd->imm()->IsFlushPending());
```
Fix this by reverting the behavior. In non-atomic-flush case, a column family can appear in the flush queue at most once at the same time.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7362
Test Plan:
make check
Also run stress test successfully for 10 times.
```
make crash_test
```
Reviewed By: ajkr
Differential Revision: D25172996
Pulled By: riversand963
fbshipit-source-id: f1559b6366cc609e961e3fc83fae548f1fad08ce
Summary:
WAL may be truncated to an incomplete record due to crash while writing
the last record or corruption. In the former case, no hole will be
produced since no ACK'd data was lost. In the latter case, a hole could
be produced without this PR since we proceeded to recover the next WAL
as if nothing happened. This PR changes the record reading code to
always report a corruption for incomplete records in
`kPointInTimeRecovery` mode, and the upper layer will only ignore them
if the next WAL has consecutive seqnum (i.e., we are guaranteed no
hole).
While this solves the hole problem for the case of incomplete
records, the possibility is still there if the WAL is corrupted by
truncation to an exact record boundary. This PR also regresses how much data
can be recovered when writes are mixed with/without
`WriteOptions::disableWAL`, as then we can not distinguish between a
seqnum gap caused by corruption and a seqnum gap caused by a `disableWAL` write.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7701
Test Plan:
Interestingly there already was a test for this case
(`DBWALTestWithParams.kPointInTimeRecovery`); it just had a typo bug in
the verification that prevented it from noticing holes in recovery.
Reviewed By: anand1976
Differential Revision: D25111765
Pulled By: ajkr
fbshipit-source-id: 5e330b13b1ee2b5be096cea9d0ff6075843e57b6
Summary:
The patch adds basic garbage collection support to the integrated BlobDB
implementation. Valid blobs residing in the oldest blob files are relocated
as they are encountered during compaction. The threshold that determines
which blob files qualify is computed based on the configuration option
`blob_garbage_collection_age_cutoff`, which was introduced in https://github.com/facebook/rocksdb/issues/7661 .
Once a blob is retrieved for the purposes of relocation, it passes through the
same logic that extracts large values to blob files in general. This means that
if, for instance, the size threshold for key-value separation (`min_blob_size`)
got changed or writing blob files got disabled altogether, it is possible for the
value to be moved back into the LSM tree. In particular, one way to re-inline
all blob values if needed would be to perform a full manual compaction with
`enable_blob_files` set to `false`, `enable_blob_garbage_collection` set to
`true`, and `blob_file_garbage_collection_age_cutoff` set to `1.0`.
Some TODOs that I plan to address in separate PRs:
1) We'll have to measure the amount of new garbage in each blob file and log
`BlobFileGarbage` entries as part of the compaction job's `VersionEdit`.
(For the time being, blob files are cleaned up solely based on the
`oldest_blob_file_number` relationships.)
2) When compression is used for blobs, the compression type hasn't changed,
and the blob still qualifies for being written to a blob file, we can simply copy
the compressed blob to the new file instead of going through decompression
and compression.
3) We need to update the formula for computing write amplification to account
for the amount of data read from blob files as part of GC.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7694
Test Plan: `make check`
Reviewed By: riversand963
Differential Revision: D25069663
Pulled By: ltamasi
fbshipit-source-id: bdfa8feb09afcf5bca3b4eba2ba72ce2f15cd06a
Summary:
This PR updates `MemTable::Add()`, `MemTable::Update()`, and
`MemTable::UpdateCallback()` to return `Status` objects, and adapts the
client code in `MemTableInserter`. The goal is to prepare these
functions for key-value checksum, where we want to verify key-value
integrity while adding to memtable. After this PR, the memtable mutation
functions can report a failed integrity check by returning `Status::Corruption`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7656
Reviewed By: riversand963
Differential Revision: D24900497
Pulled By: ajkr
fbshipit-source-id: 1a7e80581e3774676f2bbba2f0a0b04890f40009
Summary:
Allow corruption_test to run on custom env loaded via
`Env::LoadEnv()`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7699
Test Plan:
```
make corruption_test
./corruption_test
```
Also run on in-house custom env.
Reviewed By: zhichao-cao
Differential Revision: D25135525
Pulled By: riversand963
fbshipit-source-id: 7941e7ce342dc88ec2cd63e90f7674a2f57de6b7
Summary:
An empty WAL won't be backed up by the BackupEngine. So if we track the empty WALs in MANIFEST, then when restoring from a backup, it may report corruption that the empty WAL is missing, which is correct because the WAL is actually in the main DB but not in the backup DB, but missing an empty WAL does not logically break DB consistency.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7697
Test Plan: watch existing tests to pass
Reviewed By: pdillinger
Differential Revision: D25077194
Pulled By: cheng-chang
fbshipit-source-id: 01917b57234b92b6063925f2ee9452c5732bdc03
Summary:
The logic for computing min_log_number_to_keep in atomic flush was incorrect.
For example, when all column families are flushed, the min_log_number_to_keep should be the latest new log. But the incorrect logic calls `PrecomputeMinLogNumberToKeepNon2PC` for each column family, and returns the minimum of them. However, `PrecomputeMinLogNumberToKeepNon2PC(cf)` assumes column families other than `cf` are flushed, but in case all column families are flushed, this assumption is incorrect.
Without this fix, the WAL referenced by the computed min_log_number_to_keep may actually contain no unflushed data, so the WAL might have actually been deleted from disk on recovery, then an incorrect error `Corruption: missing WAL` will be reported.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7660
Test Plan:
run `make crash_test_with_atomic_flush` on devserver
added a unit test in `db_flush_test`
Reviewed By: riversand963
Differential Revision: D24906265
Pulled By: cheng-chang
fbshipit-source-id: 08deda62e71f67f59e3b7925cdd86dd09bd4f430
Summary:
In db_basic_test.cc, there are two tests that rely on the underlying
system's `LockFile` support to function correctly:
DBBasicTest.OpenWhenOpen and DBBasicTest.CheckLock. In both tests,
re-opening a db using `DB::Open` is expected to fail because the second
open cannot lock the LOCK file. Some distributed file systems, e.g. HDFS
do not support the POSIX-style file lock. Therefore, these unit tests will cause
assertion failure and the second `Open` will create a db instance.
Currently, these db instances are not closed after the assertion
failure. Since these db instances are registered with some process-wide, static
data structures, e.g. `PeriodicWorkScheduler::Default()`, they can still be
accessed after the unit tests. However, the `Env` object created for this db
instance is destroyed when the test finishes in `~DBTestBase()`. Consequently,
it causes illegal memory access.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7682
Test Plan:
Run the following on a distrubited file system:
```
make check
```
Reviewed By: anand1976
Differential Revision: D25004215
Pulled By: riversand963
fbshipit-source-id: f4327d7716c0e72b13bb43737ec9a5d156da4d52
Summary:
An application may accidentally write merge operands without properly configuring `merge_operator`. We should alert them as early as possible that there's an API misuse. Previously RocksDB only notified them when a query or background operation needed to merge but couldn't. With this PR, RocksDB notifies them of the problem before applying the merge operand to the memtable (although it may already be in WAL, which seems it'd cause a crash loop until they enable `merge_operator`).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7667
Reviewed By: riversand963
Differential Revision: D24933360
Pulled By: ajkr
fbshipit-source-id: 3a4a2ceb0b7aed184113dd03b8efd735a8332f7f
Summary:
If fsync is disabled in a unit test, then do not track WAL in MANIFEST, because on DB recovery, the WAL might be missing because the directory is not fsynced.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7669
Test Plan: Tests with fsync enabled should pass.
Reviewed By: riversand963
Differential Revision: D24941431
Pulled By: cheng-chang
fbshipit-source-id: ab3ff0f90769795cfb4e4d6dcf084ea5545d1975
Summary:
Added experimental public API for Ribbon filter:
NewExperimentalRibbonFilterPolicy(). This experimental API will
take a "Bloom equivalent" bits per key, and configure the Ribbon
filter for the same FP rate as Bloom would have but ~30% space
savings. (Note: optimize_filters_for_memory is not yet implemented
for Ribbon filter. That can be added with no effect on schema.)
Internally, the Ribbon filter is configured using a "one_in_fp_rate"
value, which is 1 over desired FP rate. For example, use 100 for 1%
FP rate. I'm expecting this will be used in the future for configuring
Bloom-like filters, as I expect people to more commonly hold constant
the filter accuracy and change the space vs. time trade-off, rather than
hold constant the space (per key) and change the accuracy vs. time
trade-off, though we might make that available.
### Benchmarking
```
$ ./filter_bench -impl=2 -quick -m_keys_total_max=200 -average_keys_per_filter=100000 -net_includes_hashing
Building...
Build avg ns/key: 34.1341
Number of filters: 1993
Total size (MB): 238.488
Reported total allocated memory (MB): 262.875
Reported internal fragmentation: 10.2255%
Bits/key stored: 10.0029
----------------------------
Mixed inside/outside queries...
Single filter net ns/op: 18.7508
Random filter net ns/op: 258.246
Average FP rate %: 0.968672
----------------------------
Done. (For more info, run with -legend or -help.)
$ ./filter_bench -impl=3 -quick -m_keys_total_max=200 -average_keys_per_filter=100000 -net_includes_hashing
Building...
Build avg ns/key: 130.851
Number of filters: 1993
Total size (MB): 168.166
Reported total allocated memory (MB): 183.211
Reported internal fragmentation: 8.94626%
Bits/key stored: 7.05341
----------------------------
Mixed inside/outside queries...
Single filter net ns/op: 58.4523
Random filter net ns/op: 363.717
Average FP rate %: 0.952978
----------------------------
Done. (For more info, run with -legend or -help.)
```
168.166 / 238.488 = 0.705 -> 29.5% space reduction
130.851 / 34.1341 = 3.83x construction time for this Ribbon filter vs. lastest Bloom filter (could make that as little as about 2.5x for less space reduction)
### Working around a hashing "flaw"
bloom_test discovered a flaw in the simple hashing applied in
StandardHasher when num_starts == 1 (num_slots == 128), showing an
excessively high FP rate. The problem is that when many entries, on the
order of number of hash bits or kCoeffBits, are associated with the same
start location, the correlation between the CoeffRow and ResultRow (for
efficiency) can lead to a solution that is "universal," or nearly so, for
entries mapping to that start location. (Normally, variance in start
location breaks the effective association between CoeffRow and
ResultRow; the same value for CoeffRow is effectively different if start
locations are different.) Without kUseSmash and with num_starts > 1 (thus
num_starts ~= num_slots), this flaw should be completely irrelevant. Even
with 10M slots, the chances of a single slot having just 16 (or more)
entries map to it--not enough to cause an FP problem, which would be local
to that slot if it happened--is 1 in millions. This spreadsheet formula
shows that: =1/(10000000*(1 - POISSON(15, 1, TRUE)))
As kUseSmash==false (the setting for Standard128RibbonBitsBuilder) is
intended for CPU efficiency of filters with many more entries/slots than
kCoeffBits, a very reasonable work-around is to disallow num_starts==1
when !kUseSmash, by making the minimum non-zero number of slots
2*kCoeffBits. This is the work-around I've applied. This also means that
the new Ribbon filter schema (Standard128RibbonBitsBuilder) is not
space-efficient for less than a few hundred entries. Because of this, I
have made it fall back on constructing a Bloom filter, under existing
schema, when that is more space efficient for small filters. (We can
change this in the future if we want.)
TODO: better unit tests for this case in ribbon_test, and probably
update StandardHasher for kUseSmash case so that it can scale nicely to
small filters.
### Other related changes
* Add Ribbon filter to stress/crash test
* Add Ribbon filter to filter_bench as -impl=3
* Add option string support, as in "filter_policy=experimental_ribbon:5.678;"
where 5.678 is the Bloom equivalent bits per key.
* Rename internal mode BloomFilterPolicy::kAuto to kAutoBloom
* Add a general BuiltinFilterBitsBuilder::CalculateNumEntry based on
binary searching CalculateSpace (inefficient), so that subclasses
(especially experimental ones) don't have to provide an efficient
implementation inverting CalculateSpace.
* Minor refactor FastLocalBloomBitsBuilder for new base class
XXH3pFilterBitsBuilder shared with new Standard128RibbonBitsBuilder,
which allows the latter to fall back on Bloom construction in some
extreme cases.
* Mostly updated bloom_test for Ribbon filter, though a test like
FullBloomTest::Schema is a next TODO to ensure schema stability
(in case this becomes production-ready schema as it is).
* Add some APIs to ribbon_impl.h for configuring Ribbon filters.
Although these are reasonably covered by bloom_test, TODO more unit
tests in ribbon_test
* Added a "tool" FindOccupancyForSuccessRate to ribbon_test to get data
for constructing the linear approximations in GetNumSlotsFor95PctSuccess.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7658
Test Plan:
Some unit tests updated but other testing is left TODO. This
is considered experimental but laying down schema compatibility as early
as possible in case it proves production-quality. Also tested in
stress/crash test.
Reviewed By: jay-zhuang
Differential Revision: D24899349
Pulled By: pdillinger
fbshipit-source-id: 9715f3e6371c959d923aea8077c9423c7a9f82b8
Summary:
https://github.com/facebook/rocksdb/issues/7556 enables `CompactionIterator` to perform garbage collection during compaction according
to a lower bound (user-defined) timestamp `full_history_ts_low_`.
This PR adds a data member `full_history_ts_low_` of type `std::string` to `FlushJob`, and
`full_history_ts_low_` does not change during flush. `FlushJob` will pass a pointer to this data member
to the `CompactionIterator` used during flush.
Also refactored flush_job_test.cc to re-use some existing code, which is actually the majority of this PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7655
Test Plan: make check
Reviewed By: ltamasi
Differential Revision: D24933340
Pulled By: riversand963
fbshipit-source-id: 2e584bfd0cf6e5c295ab1af264e68e9d6a12fca3
Summary:
https://github.com/facebook/rocksdb/pull/7461 accidentally broke
`InternalStats::DumpCFStats` by making `DumpCFFileHistogram` overwrite
the output of `DumpCFStatsNoFileHistogram` instead of appending to it,
resulting in only the file histogram related information getting logged.
The patch fixes this by reverting to appending in `DumpCFFileHistogram`.
Fixes https://github.com/facebook/rocksdb/issues/7664 .
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7666
Test Plan: Ran `make check` and checked the info log of `db_bench`.
Reviewed By: riversand963
Differential Revision: D24929051
Pulled By: ltamasi
fbshipit-source-id: 636a3d5ebb5ce23de4f3fe4f03ad3f16cb2858f8
Summary:
https://github.com/facebook/rocksdb/issues/7556 enables `CompactionIterator` to perform garbage collection during compaction according
to a lower bound (user-defined) timestamp `full_history_ts_low_`.
This PR adds a data member `full_history_ts_low_` of type `std::string` to `CompactionJob`, and
`full_history_ts_low_` does not change during compaction. `CompactionJob` will pass a pointer to this
data member to the `CompactionIterator` used during compaction.
Also refactored compaction_job_test.cc to re-use some existing code, which is actually the majority of this PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7657
Test Plan: make check
Reviewed By: ltamasi
Differential Revision: D24913803
Pulled By: riversand963
fbshipit-source-id: 11ad5329ddac365667152e7b3b02f84182c0ca8e
Summary:
`CompactionProxy` is currently both a concrete class used for actual `Compaction`s
and a base class that `FakeCompaction` (which is used in `compaction_iterator_test`)
is derived from. This is bad from an OO design standpoint, and also results in
`FakeCompaction` containing an (uninitialized and unused) `Compaction*` member.
The patch fixes this by making `CompactionProxy` a pure interface and introducing
a separate concrete class `RealCompaction` for non-test/non-fake compactions. It
also removes an unused parameter from the virtual method `level`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7662
Test Plan: `make check`
Reviewed By: riversand963
Differential Revision: D24907680
Pulled By: ltamasi
fbshipit-source-id: c100ecb1beef4b0ada35e799116c5bda71719ee7
Summary:
Previously, even when `bottommost_compression_opts`'s `enabled` flag was set, it only took effect when
`bottommost_compression` was also set to something other than `kDisableCompressionOption`.
This wasn't documented and, if we kept the old behavior, it'd make
things complicated like the migration instructions in https://github.com/facebook/rocksdb/issues/7619. We can
simplify the API by making `bottommost_compression_opts` always take
effect when its `enabled` flag is set.
Fixes https://github.com/facebook/rocksdb/issues/7631.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7633
Reviewed By: ltamasi
Differential Revision: D24710358
Pulled By: ajkr
fbshipit-source-id: bbbdf9c1b53c63a4239d902cc3f5a11da1874647
Summary:
Added a few classes in the same class hierarchy to remove code duplication and
refactor the logic of reading and processing MANIFEST files.
New classes are as follows.
```
class VersionEditHandlerBase;
class ListColumnFamiliesHandler : VersionEditHandlerBase;
class FileChecksumRetriever : VersionEditHandlerBase;
class DumpManifestHandler : VersionEditHandler;
```
Classes that already existed before this PR are as follows.
```
class VersionEditHandler : VersionEditHandlerBase;
```
With these classes, refactored functions: `VersionSet::Recover()`,
`VersionSet::ListColumnFamilies()`, `VersionSet::DumpManifest()`,
`GetFileChecksumFromManifest()`.
Test Plan (devserver):
```
make check
COMPILE_WITH_ASAN=1 make check
```
These refactored code, especially recovery-related logic, will be tested intensively by
all existing unit tests and stress tests. For example, run
```
make crash_test
```
Verified 3 successful runs on devserver.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6581
Reviewed By: ajkr
Differential Revision: D20616217
Pulled By: riversand963
fbshipit-source-id: 048c7743aa4be2623ccd0cc3e61c0027e604e78b
Summary:
Since the hashes should not be persisted in output_validator
nor mock_env.
Also updated NPHash64 to use 64-bit seed, and comments.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7632
Test Plan:
make check, and new build setting that enables modification
to NPHash64, to check for behavior depending on specific values. Added
that setting to one of the CircleCI configurations.
Reviewed By: jay-zhuang
Differential Revision: D24833780
Pulled By: pdillinger
fbshipit-source-id: 02a57652ccf1ac105fbca79e77875bb7bf7c071f
Summary:
There is an undocumented behavior about a certain combination of options and operations.
- inplace_update_support = true, and
- call `SeekForPrev()`, `SeekToLast()`, and/or `Prev()` on unflushed data.
We should stop the backward iteration and report an error of `Status::NotSupported`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7618
Test Plan: make check
Reviewed By: pdillinger
Differential Revision: D24769619
Pulled By: riversand963
fbshipit-source-id: 81d199fa55ed4739ab10e719cc345a992238ccbb
Summary:
https://github.com/facebook/rocksdb/issues/7556 introduced support for compaction iterator to perform timestamp-aware garbage collection.
However, there was a bug. The comparison between `ikey_.user_key` and `current_user_key_` should happen
before `key_ = current_key_.SetInternalKey(key_, &ikey_);` (line 336 of compaction_iterator.cc).
Otherwise, after this line, `current_key_` is always the same as `ikey_.user_key`.
This PR also re-arranged the order of some data members because some of them are state variables of `CompactionIterator` while others are inputs from callers.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7645
Test Plan: make check
Reviewed By: ltamasi
Differential Revision: D24845028
Pulled By: riversand963
fbshipit-source-id: c7e79914832701462b86867e8463cd463b6c0c25
Summary:
After replaying the WALs, the memtables are flushed synchronously to L0 instead of being flushed in background. Currently, we only track WAL obsoletion events in the code path of background flush jobs. This PR tracks these events in RecoverLogFiles.
After this change, we can enable `track_and_verify_wal_in_manifest` in `db_stress`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7649
Test Plan: `python tools/db_crashtest.py whitebox`
Reviewed By: riversand963
Differential Revision: D24824501
Pulled By: cheng-chang
fbshipit-source-id: 207129f7b845c50b333680ce6818a68a2fad54b9
Summary:
Consider the following sequence of events:
1. Db flushed an SST with file number N, appended to MANIFEST, and tried to sync the MANIFEST.
2. Syncing MANIFEST failed and db crashed.
3. Db tried to recover with this MANIFEST. In the meantime, no entry about the newly-flushed SST was found in the MANIFEST. Therefore, RocksDB replayed WAL and tried to flush to an SST file reusing the same file number N. This failed because file system does not support overwrite. Then Db deleted this file.
4. Db crashed again.
5. Db tried to recover. When db read the MANIFEST, there was an entry referencing N.sst. This could happen probably because the append in step 1 finally reached the MANIFEST and became visible. Since N.sst had been deleted in step 3, recovery failed.
It is possible that N.sst created in step 1 is valid. Although step 3 would still fail since the MANIFEST was not synced properly in step 1 and 2, deleting N.sst would make it impossible for the db to recover even if the remaining part of MANIFEST was appended and visible after step 5.
After this PR, in step 3, immediately after recovering from MANIFEST, a new MANIFEST is created, then we find that N.sst is not referenced in the MANIFEST, so we delete it, and we'll not reuse N as file number. Then in step 5, since the new MANIFEST does not contain N.sst, the recovery failure situation in step 5 won't happen.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7621
Test Plan:
1. some tests are updated, because these tests assume that new MANIFEST is created after WAL recovery.
2. a new unit test is added in db_basic_test to simulate step 3.
Reviewed By: riversand963
Differential Revision: D24668144
Pulled By: cheng-chang
fbshipit-source-id: 90d7487fbad2bc3714f5ede46ea949895b15ae3b
Summary:
When a WAL is synced, an edit is written to MANIFEST.
After flushing memtables, the obsoleted WALs are piggybacked to MANIFEST while writing the new L0 files to MANIFEST.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7601
Test Plan:
`track_and_verify_wals_in_manifest` is enabled by default for all tests extending `DBBasicTest`, and in db_stress_test.
Unit test `wal_edit_test`, `version_edit_test`, and `version_set_test` are also updated.
Watch all tests to pass.
Reviewed By: ltamasi
Differential Revision: D24553957
Pulled By: cheng-chang
fbshipit-source-id: 66a569ff1bdced38e22900bd240b73113906e040
