If log recycling is enabled with the rocksdb (recycle_log_file_num=16)
db->Writebatch is erroring out with keynotfound after ~5-6 hours of run
(1M seq but can happen to any workload I guess).See my detailed bug
report here (https://github.com/facebook/rocksdb/issues/1303).
This commit is the fix for this, a check is been added not to delete
the log file if it is already there in the recycle list.
Test Plan:
Unit tested it and ran the similar profile. Not reproducing anymore.
Summary: As title, make sure Prev() works as expected with Next() when the current iter->key() in the range of the same prefix in prefix seek mode
Test Plan: make all check -j64 (add prefix_test with PrefixSeekModePrev test case)
Reviewers: andrewkr, sdong, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: yoshinorim, andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D61419
Summary:
After 1b8a2e8fdd, DB Pointer is passed to WriteBatchInternal::InsertInto() while DB recovery. This can cause deadlock if options.max_successive_merges hits. In that case DB::Get() will be called. Get() will try to acquire the DB mutex, which is already held by the DB::Open(), causing a deadlock condition.
This commit mitigates the problem by not passing the DB pointer unless 2PC is allowed.
Test Plan: Add a new test and run it.
Reviewers: IslamAbdelRahman, andrewkr, kradhakrishnan, horuff
Reviewed By: kradhakrishnan
Subscribers: leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D62625
Summary:
Move the manual memtable flush for databases containing data that has
bypassed the WAL from DBImpl's destructor to CancleAllBackgroundWork().
CancelAllBackgroundWork() is a publicly exposed API which allows
async operations performed by background threads to be disabled on a
database. In effect, this places the database into a "shutdown" state
in advance of calling the database object's destructor. No compactions
or flushing of SST files can occur once a call to this API completes.
When writes are issued to a database with WriteOptions::disableWAL
set to true, DBImpl::has_unpersisted_data_ is set so that
memtables can be flushed when the database object is destroyed. If
CancelAllBackgroundWork() has been called prior to DBImpl's destructor,
this flush operation is not possible and is skipped, causing unnecessary
loss of data.
Since CancelAllBackgroundWork() is already invoked by DBImpl's destructor
in order to perform the thread join portion of its cleanup processing,
moving the manual memtable flush to CancelAllBackgroundWork() ensures
data is persisted regardless of client behavior.
Test Plan:
Write an amount of data that will not cause a memtable flush to a rocksdb
database with all writes marked with WriteOptions::disableWAL. Properly
"close" the database. Reopen database and verify that the data was
persisted.
Reviewers: IslamAbdelRahman, yiwu, yoshinorim, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D62277
Summary: With read_options.background_purge_on_iterator_cleanup=true, File deletion and closing can still happen in forward iterator, or WAL file closing. Cover those cases too.
Test Plan: I am adding unit tests.
Reviewers: andrewkr, IslamAbdelRahman, yiwu
Reviewed By: yiwu
Subscribers: leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D61503
Summary:
My understanding is that the purpose of write stall triggers are to wait for auto-compaction to catch up. Without auto-compaction, we don't need to stall writes.
Also with this diff, flush/compaction conditions are recalculated on dynamic option change. Previously the conditions are recalculate only when write stall options are changed.
Test Plan: See the new test. Removed two tests that are no longer valid.
Reviewers: IslamAbdelRahman, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D61437
Summary:
Stale log files can be deleted out of order. This can happen for various reasons. One of the reason is that no data is ever inserted to a column family and we have an optimization to update its log number, but not all the old log files are cleaned up (the case shown in the unit tests added). It can also happen when we simply delete multiple log files out of order.
This causes data corruption because we simply increase seqID after processing the next row and we may end up with writing data with smaller seqID than what is already flushed to memtables.
In DB recovery, for the oldest files we are replaying, if there it contains no data for any column family, we ignore the sequence IDs in the file.
Test Plan: Add two unit tests that fail without the fix.
Reviewers: IslamAbdelRahman, igor, yiwu
Reviewed By: yiwu
Subscribers: hermanlee4, yoshinorim, leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D60891
Summary: Multiput atomiciy is broken across multiple column families if we don't sync WAL before flushing one column family. The WAL file may contain a write batch containing writes to a key to the CF to be flushed and a key to other CF. If we don't sync WAL before flushing, if machine crashes after flushing, the write batch will only be partial recovered. Data to other CFs are lost.
Test Plan: Add a new unit test which will fail without the diff.
Reviewers: yhchiang, IslamAbdelRahman, igor, yiwu
Reviewed By: yiwu
Subscribers: yiwu, leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D60915
Summary:
add DestroyColumnFamilyHandle(ColumnFamilyHandle**) to close column family instead of deleting cfh*
User should call this to close a cf and then we can detect the deletion in this function.
Test Plan: make all check -j64
Reviewers: andrewkr, yiwu, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D60765
Summary:
When write stalls because of auto compaction is disabled, or stop write trigger is reached,
user may change these two options to unblock writes. Unfortunately we had issue where the write
thread will block the attempt to persist the options, thus creating a deadlock. This diff
fix the issue and add two test cases to detect such deadlock.
Test Plan:
Run unit tests.
Also, revert db_impl.cc to master (but don't revert `DBImpl::BackgroundCompaction:Finish` sync point) and run db_options_test. Both tests should hit deadlock.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D60627
Summary:
DB::AddFile(std::string file_path) API that allow them to ingest an SST file created using SstFileWriter
We want to update this interface to be able to accept a list of files that will be ingested, DB::AddFile(std::vector<std::string> file_path_list).
Test Plan:
Add test case `AddExternalSstFileList` in `DBSSTTest`. To make sure:
1. files key ranges are not overlapping with each other
2. each file key range dont overlap with the DB key range
3. make sure no snapshots are held
Reviewers: andrewkr, sdong, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D58587
Summary: We saw instances where total_log_size is off the real value, but I'm not able to reproduce it. Add more logging to help debugging when it happens again.
Test Plan: Run the unit test and see the logging.
Reviewers: andrewkr, yhchiang, igor, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D60081
Summary: Add option write_buffer_manager to help users control total memory spent on memtables across multiple DB instances.
Test Plan: Add a new unit test.
Reviewers: yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: adela, benj, sumeet, muthu, leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D59925
Summary: Reported in T11889874. When registering the cleanup function we should copy the option so that we can still access it if ReadOptions is deleted.
Test Plan: Add a unit test to reproduce this bug.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D60087
Summary:
Add a read option `background_purge_on_iterator_cleanup` to avoid deleting files in foreground when destroying iterators.
Instead, a job is scheduled in high priority queue and would be executed in a separate background thread.
Test Plan: Add a variant of PurgeObsoleteFileTest. Turn on background purge option in the new test, and use sleeping task to ensure files are deleted in background.
Reviewers: IslamAbdelRahman, sdong
Reviewed By: IslamAbdelRahman
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D59499
Summary:
DB::AddFile() right now always add the ingested file to L0
update the logic to add the file to the lowest possible level
Test Plan: unit tests
Reviewers: jkedgar, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D59637
Summary: filter_deltes is not a frequently used feature. Remove it.
Test Plan: Run all test suites.
Reviewers: igor, yhchiang, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D59427
Summary:
We dont report the bytes that we ingested from AddFile which make the write amplification numbers incorrect
Update InternalStats and add logging for AddFile()
Test Plan: Make sure the code compile and existing tests pass
Reviewers: lightmark, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D59763
Summary:
Add option to not flush memtable on open()
In case the option is enabled, don't delete existing log files by not updating log numbers to MANIFEST.
Will still flush if we need to (e.g. memtable full in the middle). In that case we also flush final memtable.
If wal_recovery_mode = kPointInTimeRecovery, do not halt immediately after encounter corruption. Instead, check if seq id of next log file is last_log_sequence + 1. In that case we continue recovery.
Test Plan: See unit test.
Reviewers: dhruba, horuff, sdong
Reviewed By: sdong
Subscribers: benj, yhchiang, andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D57813
Summary: Backup options file to private directory
Test Plan:
backupable_db_test.cc, BackupOptions
Modify DB options by calling OpenDB for 3 times. Check the latest options file is in the right place. Also check no redundent files are backuped.
Reviewers: andrewkr
Reviewed By: andrewkr
Subscribers: leveldb, dhruba, andrewkr
Differential Revision: https://reviews.facebook.net/D59373
Summary:
MemTableList::current_ could be written by background flush thread and
simultaneously read in the user thread (NumNotFlushed() is used in
SwitchMemtable()). Use the lock to prevent this case. Found the error from tsan.
Related: D58833
Test Plan:
$ OPT=-g COMPILE_WITH_TSAN=1 make -j64 db_test
$ TEST_TMPDIR=/dev/shm/rocksdb ./db_test --gtest_filter=DBTest.RepeatedWritesToSameKey
Reviewers: lightmark, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D59139
* Create a callback for memtable becoming immutable
Create a callback for memtable becoming immutable
Create a callback for memtable becoming immutable
moved notification outside the lock
Move sealed notification to unlocked portion of SwitchMemtable
* fix lite build
Summary:
We see some write stalls because of number of unflushed memtables. With existing logging I couldn't figure out what's happening exactly. See internal task t11446054 for details if interested. This diff adds:
- logging of memtable creation at info level; I wanted it on multiple occasions for different reasons; also include number of immutable memtables,
- logging of number of remaining immutable memtables after a flush.
Test Plan: ran tests
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D58833
Summary: Make CompactionOptionsFIFO a part of mutable_cf_options
Test Plan: UT
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, lgalanis, dhruba
Differential Revision: https://reviews.facebook.net/D58653
kPointInTimeRecovery is indistinguishable from
kTolerateCorruptedTailRecords in recycle mode since we define
the "end" of the log as the first corrupt record we encounter.
kAbsoluteConsistency doesn't make sense because even a clean
shutdown leaves old junk at the end of the log file.
Signed-off-by: Sage Weil <sage@redhat.com>
Summary:
NotifyOnCompactionCompleted can unlock the mutex.
That mean that we can schedule a background compaction that will start before we ReleaseCompactionFiles().
Test Plan:
added unittest
existing unittest
Reviewers: yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D58065
Summary:
- Make sure we clean up recovered_transactions_ on DBImpl destructor
- delete leaked txns and env in TransactionTest
Test Plan: Run transaction_test under valgrind
Reviewers: sdong, andrewkr, yhchiang, horuff
Reviewed By: horuff
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D58263
Summary: DBTestXactLogIterator.TransactionLogIterator was failing due the sequence gaps. This was caused by an off-by-one error when calculating the new sequence number after recovering from logs.
Test Plan: db_log_iter_test
Reviewers: andrewkr
Subscribers: andrewkr, hermanlee4, dhruba, IslamAbdelRahman
Differential Revision: https://reviews.facebook.net/D58053
Summary:
1. prepare()
2. crash
3. recover
4. commit()
5. crash
6. data is lost
This is due to the transaction data still only residing in the WAL but because the logs were flushed on the first recovery the data is ignored on the second recovery. We must scan all logs found on recovery and only ignore redundant data at the time of replay. It is not possible to know which logs still contain relevant data at time of recovery. We cannot simply ignore a log because all of the non-2pc data it contains has already been written to L0.
The changes made to MemTableInserter are to ensure that prepared sections are still recovered even if all of the non-2pc data in that log has already been flushed to L0.
Test Plan: Provided test.
Reviewers: sdong
Subscribers: andrewkr, hermanlee4, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D57729
Summary:
Consider the following WAL with 4 batch entries prefixed with their sequence at time of memtable insert.
[1: BEGIN_PREPARE, PUT, PUT, PUT, PUT, END_PREPARE(a)]
[1: BEGIN_PREPARE, PUT, PUT, PUT, PUT, END_PREPARE(b)]
[4: COMMIT(a)]
[7: COMMIT(b)]
The first two batches do not consume any sequence numbers so are both prefixed with seq=1.
For 2pc commit, memtable insertion takes place before COMMIT batch is written to WAL.
We can see that sequence number consumption takes place between WAL entries giving us the seemingly sparse sequence prefix for WAL entries.
This is a valid WAL.
Because with 2PC markers one WriteBatch points to another batch containing its inserts a writebatch can consume more or less sequence numbers than the number of sequence consuming entries that it contains.
We can see that, given the entries in the WAL, 6 sequence ids were consumed. Yet on recovery the maximum sequence consumed would be 7 + 3 (the number of sequence numbers consumed by COMMIT(b))
So, now upon recovery we must track the actual consumption of sequence numbers.
In the provided scenario there will be no sequence gaps, but it is possible to produce a sequence gap. This should not be a problem though. correct?
Test Plan: provided test.
Reviewers: sdong
Subscribers: andrewkr, leveldb, dhruba, hermanlee4
Differential Revision: https://reviews.facebook.net/D57645
Summary:
This diff is built on top of WriteBatch modification: https://reviews.facebook.net/D54093 and adds the required functionality to rocksdb core necessary for rocksdb to support 2PC.
modfication of DBImpl::WriteImpl()
- added two arguments *uint64_t log_used = nullptr, uint64_t log_ref = 0;
- *log_used is an output argument which will return the log number which the incoming batch was inserted into, 0 if no WAL insert took place.
- log_ref is a supplied log_number which all memtables inserted into will reference after the batch insert takes place. This number will reside in 'FindMinPrepLogReferencedByMemTable()' until all Memtables insertinto have flushed.
- Recovery/writepath is now aware of prepared batches and commit and rollback markers.
Test Plan: There is currently no test on this diff. All testing of this functionality takes place in the Transaction layer/diff but I will add some testing.
Reviewers: IslamAbdelRahman, sdong
Subscribers: leveldb, santoshb, andrewkr, vasilep, dhruba, hermanlee4
Differential Revision: https://reviews.facebook.net/D56919
Summary:
Add a new option that can be used to set a specific compression algorithm for bottommost level.
This option will only affect levels larger than base level.
I have also updated CompactionJobInfo to include the compression algorithm used in compaction
Test Plan:
added new unittest
existing unittests
Reviewers: andrewkr, yhchiang, sdong
Reviewed By: sdong
Subscribers: lightmark, andrewkr, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D57669
Summary: Added EventListener::OnTableFileCreationStarted. EventListener::OnTableFileCreated will be called on failure case. User can check creation status via TableFileCreationInfo::status.
Test Plan: unit test.
Reviewers: dhruba, yhchiang, ott, sdong
Reviewed By: sdong
Subscribers: sdong, kradhakrishnan, IslamAbdelRahman, andrewkr, yhchiang, leveldb, ott, dhruba
Differential Revision: https://reviews.facebook.net/D56337
Summary: Warning is printed out with USE_CLANG=1 when including jemalloc.h. Disable it in that case.
Test Plan: Run db_bench with USE_CLANG=1 and not. Make sure they can all build and jemalloc status is printed out in the case where USE_CLANG is not set.
Reviewers: andrewkr, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57399
Summary: It looks like we mistakenly enable JEMALLOC even if it's not available on the machine, that's why travis is failing
Test Plan:
check on my devserver
check on my mac
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57345
Summary:
Introduced option to dump malloc statistics using new option flag.
Added new command line option to db_bench tool to enable this
funtionality.
Also extended build to support environments with/without jemalloc.
Test Plan:
1) Build rocksdb using `make` command. Launch the following command
`./db_bench --benchmarks=fillrandom --dump_malloc_stats=true
--num=10000000` end verified that jemalloc dump is present in LOG file.
2) Build rocksdb using `DISABLE_JEMALLOC=1 make db_bench -j32` and ran
the same db_bench tool and found the following message in LOG file:
"Please compile with jemalloc to enable malloc dump".
3) Also built rocksdb using `make` command on MacOS to verify behavior
in non-FB environment.
Also to debug build configuration change temporary changed
AM_DEFAULT_VERBOSITY = 1 in Makefile to see compiler and build
tools output. For case 1) -DROCKSDB_JEMALLOC was present in compiler
command line. For both 2) and 3) this flag was not present.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57321
Summary: CompactedDB skips memtable. So we shouldn't use compacted DB if there is outstanding WAL files.
Test Plan: Change to options.max_open_files = -1 perf context test to create a compacted DB, which we shouldn't do.
Reviewers: yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D57057
Summary:
In https://reviews.facebook.net/D56271, we fixed an issue where
we consider flush as compaction. However, that makes us mistakenly
count FLUSH_WRITE_BYTES twice (one in flush_job and one in db_impl.)
This patch removes the one incremented in db_impl.
Test Plan: db_test
Reviewers: yiwu, andrewkr, IslamAbdelRahman, kradhakrishnan, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D57111
Summary: It is useful to print out IO stats in flush jobs too. Extend options.compaction_measure_io_stats to flush jobs and raname it.
Test Plan: Try db_bench and see the stats are printed out.
Reviewers: yhchiang
Reviewed By: yhchiang
Subscribers: kradhakrishnan, yiwu, IslamAbdelRahman, leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56769
Summary: The code assumes that if use_mmap_reads is on then use_os_buffer is also on. This make sense as by using memory mapped files for reading you are expecting the OS to cache what it needs. Add code to make sure the user does not turn off use_os_buffer when they turn on use_mmap_reads
Test Plan: New test: DBTest.MMapAndBufferOptions
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56397
Summary:
Added the column family name to the properties block. This property
is omitted only if the property is unavailable, such as when RepairDB()
writes SST files.
In a next diff, I will change RepairDB to use this new property for
deciding to which column family an existing SST file belongs. If this
property is missing, it will add it to the "unknown" column family (same
as its existing behavior).
Test Plan:
New unit test:
$ ./db_table_properties_test --gtest_filter=DBTablePropertiesTest.GetColumnFamilyNameProperty
Reviewers: IslamAbdelRahman, yhchiang, sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D55605
Summary:
When a block based table file is opened, if prefetch_index_and_filter is true, it will prefetch the index and filter blocks, putting them into the block cache.
What this feature adds: when a L0 block based table file is opened, if pin_l0_filter_and_index_blocks_in_cache is true in the options (and prefetch_index_and_filter is true), then the filter and index blocks aren't released back to the block cache at the end of BlockBasedTableReader::Open(). Instead the table reader takes ownership of them, hence pinning them, ie. the LRU cache will never push them out. Meanwhile in the table reader, further accesses will not hit the block cache, thus avoiding lock contention.
Test Plan:
'export TEST_TMPDIR=/dev/shm/ && DISABLE_JEMALLOC=1 OPT=-g make all valgrind_check -j32' is OK.
I didn't run the Java tests, I don't have Java set up on my devserver.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56133
record.size() should not be less than 12.
This "magic number" seems to be the WriteBatch header (8 byte sequence
and 4 byte count). Replaced all the places where "12" was used
by WriteBatchInternal::kHeader.
somnathr