Summary:
Throughput: 46k tps in our sysbench settings (filling the details later)
The idea is to have the simplest change that gives us a reasonable boost
in 2PC throughput.
Major design changes:
1. The WAL file internal buffer is not flushed after each write. Instead
it is flushed before critical operations (WAL copy via fs) or when
FlushWAL is called by MySQL. Flushing the WAL buffer is also protected
via mutex_.
2. Use two sequence numbers: last seq, and last seq for write. Last seq
is the last visible sequence number for reads. Last seq for write is the
next sequence number that should be used to write to WAL/memtable. This
allows to have a memtable write be in parallel to WAL writes.
3. BatchGroup is not used for writes. This means that we can have
parallel writers which changes a major assumption in the code base. To
accommodate for that i) allow only 1 WriteImpl that intends to write to
memtable via mem_mutex_--which is fine since in 2PC almost all of the memtable writes
come via group commit phase which is serial anyway, ii) make all the
parts in the code base that assumed to be the only writer (via
EnterUnbatched) to also acquire mem_mutex_, iii) stat updates are
protected via a stat_mutex_.
Note: the first commit has the approach figured out but is not clean.
Submitting the PR anyway to get the early feedback on the approach. If
we are ok with the approach I will go ahead with this updates:
0) Rebase with Yi's pipelining changes
1) Currently batching is disabled by default to make sure that it will be
consistent with all unit tests. Will make this optional via a config.
2) A couple of unit tests are disabled. They need to be updated with the
serial commit of 2PC taken into account.
3) Replacing BatchGroup with mem_mutex_ got a bit ugly as it requires
releasing mutex_ beforehand (the same way EnterUnbatched does). This
needs to be cleaned up.
Closes https://github.com/facebook/rocksdb/pull/2345
Differential Revision: D5210732
Pulled By: maysamyabandeh
fbshipit-source-id: 78653bd95a35cd1e831e555e0e57bdfd695355a4
Summary:
CreateColumnFamily() releases DB mutex after adding column family to the set and install super version (to write option file), so if users call GetAggregatedIntProperty() in the middle, then super version will be null and the process will crash. Fix it by skipping those column families without super version installed.
Maybe we should also fix the problem of releasing the lock when reading option file, but it is more risky. so I'm doing a quick and safer fix and we can investigate it later.
Closes https://github.com/facebook/rocksdb/pull/2475
Differential Revision: D5298053
Pulled By: siying
fbshipit-source-id: 4b3c8f91c60400b163fcc6cda8a0c77723be0ef6
Summary:
If ReadOptions.low_pri=true and compaction is behind, the write will either return immediate or be slowed down based on ReadOptions.no_slowdown.
Closes https://github.com/facebook/rocksdb/pull/2369
Differential Revision: D5127619
Pulled By: siying
fbshipit-source-id: d30e1cff515890af0eff32dfb869d2e4c9545eb0
Summary:
This fixes a compilation failure on Linux when the system libc is not
glibc. jemalloc's configure script incorrectly assumes that glibc is
always used on Linux systems, producing glibc-style signatures; when
the system libc is e.g. musl, the following error is observed:
```
[ 0%] Building CXX object CMakeFiles/rocksdb.dir/db/db_impl.cc.o
In file included from /go/src/github.com/cockroachdb/cockroach/c-deps/rocksdb.src/table/block.h:19:0,
from /go/src/github.com/cockroachdb/cockroach/c-deps/rocksdb.src/db/db_impl.cc:77:
/x-tools/x86_64-unknown-linux-musl/x86_64-unknown-linux-musl/sysroot/usr/include/malloc.h:19:8: error: declaration of 'size_t malloc_usable_size(void*)' has a different exception specifier
size_t malloc_usable_size(void *);
^~~~~~~~~~~~~~~~~~
In file included from /go/src/github.com/cockroachdb/cockroach/c-deps/rocksdb.src/db/db_impl.cc:20:0:
/go/native/x86_64-unknown-linux-musl/jemalloc/include/jemalloc/jemalloc.h:78:33: note: from previous declaration 'size_t malloc_usable_size(void*) throw ()'
# define je_malloc_usable_size malloc_usable_size
^
/go/native/x86_64-unknown-linux-musl/jemalloc/include/jemalloc/jemalloc.h:239:41: note: in expansion of macro 'je_malloc_usable_size'
JEMALLOC_EXPORT size_t JEMALLOC_NOTHROW je_malloc_usable_size(
^~~~~~~~~~~~~~~~~~~~~
CMakeFiles/rocksdb.dir/build.make:350: recipe for target 'CMakeFiles/rocksdb.dir/db/db_impl.cc.o' failed
```
This works around the issue by rearranging the sources such that
jemalloc's headers are never in the same scope as the system's malloc
header. The jemalloc issue has been reported as well, see:
https://github.com/jemalloc/jemalloc/issues/778.
cc tschottdorf
Closes https://github.com/facebook/rocksdb/pull/2188
Differential Revision: D5163048
Pulled By: siying
fbshipit-source-id: c553125458892def175c1be5682b0330d80b2a0d
Summary:
PipelineWriteImpl is an alternative approach to WriteImpl. In WriteImpl, only one thread is allow to write at the same time. This thread will do both WAL and memtable writes for all write threads in the write group. Pending writers wait in queue until the current writer finishes. In the pipeline write approach, two queue is maintained: one WAL writer queue and one memtable writer queue. All writers (regardless of whether they need to write WAL) will still need to first join the WAL writer queue, and after the house keeping work and WAL writing, they will need to join memtable writer queue if needed. The benefit of this approach is that
1. Writers without memtable writes (e.g. the prepare phase of two phase commit) can exit write thread once WAL write is finish. They don't need to wait for memtable writes in case of group commit.
2. Pending writers only need to wait for previous WAL writer finish to be able to join the write thread, instead of wait also for previous memtable writes.
Merging #2056 and #2058 into this PR.
Closes https://github.com/facebook/rocksdb/pull/2286
Differential Revision: D5054606
Pulled By: yiwu-arbug
fbshipit-source-id: ee5b11efd19d3e39d6b7210937b11cefdd4d1c8d
Summary:
First cut for early review; there are few conceptual points to answer and some code structure issues.
For conceptual points -
- restriction-wise, we're going to disallow ingest_behind if (use_seqno_zero_out=true || disable_auto_compaction=false), the user is responsible to properly open and close DB with required params
- we wanted to ingest into reserved bottom most level. Should we fail fast if bottom level isn't empty, or should we attempt to ingest if file fits there key-ranges-wise?
- Modifying AssignLevelForIngestedFile seems the place we we'd handle that.
On code structure - going to refactor GenerateAndAddExternalFile call in the test class to allow passing instance of IngestionOptions, that's just going to incur lots of changes at callsites.
Closes https://github.com/facebook/rocksdb/pull/2144
Differential Revision: D4873732
Pulled By: lightmark
fbshipit-source-id: 81cb698106b68ef8797f564453651d50900e153a
Summary:
Adding DB::CreateColumnFamilie() and DB::DropColumnFamilies() to bulk create/drop column families. This is to address the problem creating/dropping 1k column families takes minutes. The bottleneck is we persist options files for every single column family create/drop, and it parses the persisted options file for verification, which take a lot CPU time.
The new APIs simply create/drop column families individually, and persist options file once at the end. This improves create 1k column families to within ~0.1s. Further improvement can be merge manifest write to one IO.
Closes https://github.com/facebook/rocksdb/pull/2248
Differential Revision: D5001578
Pulled By: yiwu-arbug
fbshipit-source-id: d4e00bda671451e0b314c13e12ad194b1704aa03
Summary:
Makes max_open_files db option dynamically set-able by SetDBOptions. During the call of SetDBOptions we call SetCapacity on the table cache, which is a LRUCache.
Closes https://github.com/facebook/rocksdb/pull/2185
Differential Revision: D4979189
Pulled By: yiwu-arbug
fbshipit-source-id: ca7e8dc5e3619c79434f579be4847c0f7e56afda
Summary:
Support buck load with universal compaction.
More test cases to be added.
Closes https://github.com/facebook/rocksdb/pull/2202
Differential Revision: D4935360
Pulled By: lightmark
fbshipit-source-id: cc3ca1b6f42faa503207dab1408d6bcf393ee5b5
Summary:
Add a function to allow users to reset internal stats without restarting the DB.
Closes https://github.com/facebook/rocksdb/pull/2167
Differential Revision: D4907939
Pulled By: siying
fbshipit-source-id: ab2dd85b88aabe9380da7485320a1d460d3e1f68
Summary:
Currently level histogram is only printed out for DB stats and for default CF. This is confusing. Change to print for every CF instead.
Closes https://github.com/facebook/rocksdb/pull/2126
Differential Revision: D4865373
Pulled By: siying
fbshipit-source-id: 1c853e0ac66e00120ee931cabc9daf69ccc2d577
Summary:
DBIter, and in-turn NewDBIterator and NewArenaWrappedDBIterator, take a bunch of params. They can be reduced by passing in ReadOptions directly instead of passing in every new param separately. It also seems much cleaner as a bunch of the params towards the end seem to be optional.
(Recently I introduced max_skippable_internal_keys, which added one more to the already huge count).
Idea courtesy IslamAbdelRahman
Closes https://github.com/facebook/rocksdb/pull/2116
Differential Revision: D4857128
Pulled By: sagar0
fbshipit-source-id: 7d239df094b94bd9ea79d145cdf825478ac037a8
Summary:
siying this is a resubmission of #2081 with the 4th commit fixed. From that commit message:
> Note that the previous use of quotes in PLATFORM_{CC,CXX}FLAGS was
incorrect and caused GCC to produce the incorrect define:
>
> #define ROCKSDB_JEMALLOC -DJEMALLOC_NO_DEMANGLE 1
>
> This was the cause of the Linux build failure on the previous version
of this change.
I've tested this locally, and the Linux build succeeds now.
Closes https://github.com/facebook/rocksdb/pull/2097
Differential Revision: D4839964
Pulled By: siying
fbshipit-source-id: cc51322
Summary:
Move some files under util/ to new directories env/, monitoring/ options/ and cache/
Closes https://github.com/facebook/rocksdb/pull/2090
Differential Revision: D4833681
Pulled By: siying
fbshipit-source-id: 2fd8bef
Summary:
db_impl.cc is too large to manage. Divide db_impl.cc into db/db_impl.cc, db/db_impl_compaction_flush.cc, db/db_impl_files.cc, db/db_impl_open.cc and db/db_impl_write.cc.
Closes https://github.com/facebook/rocksdb/pull/2095
Differential Revision: D4838188
Pulled By: siying
fbshipit-source-id: c5f3059
Summary:
Refactor WriteImpl() so when I plug-in the pipeline write code (which is
an alternative approach for WriteThread), some of the logic can be
reuse. I split out the following methods from WriteImpl():
* PreprocessWrite()
* HandleWALFull() (previous MaybeFlushColumnFamilies())
* HandleWriteBufferFull()
* WriteToWAL()
Also adding a constructor to WriteThread::Writer, and move WriteContext into db_impl.h.
No real logic change in this patch.
Closes https://github.com/facebook/rocksdb/pull/2042
Differential Revision: D4781014
Pulled By: yiwu-arbug
fbshipit-source-id: d45ca18
Summary:
It is confusing to have auto_roll_logger to stay under db/, which has nothing to do with database. Move filename together as it is a dependency.
Closes https://github.com/facebook/rocksdb/pull/2080
Differential Revision: D4821141
Pulled By: siying
fbshipit-source-id: ca7d768
Summary:
`uint` is nonstandard and not a built-in type on all compilers; replace it
with the always-valid `unsigned int`. I assume this went unnoticed because
it's inside an `#ifdef ROCKDB_JEMALLOC`.
Closes https://github.com/facebook/rocksdb/pull/2075
Differential Revision: D4820427
Pulled By: ajkr
fbshipit-source-id: 0876561
Summary:
Operations like Seek/Next/Prev sometimes take too long to complete when there are many internal keys to be skipped. Adding an option, max_skippable_internal_keys -- which could be used to set a threshold for the maximum number of keys that can be skipped, will help to address these cases where it is much better to fail a request (as incomplete) than to wait for a considerable time for the request to complete.
This feature -- to fail an iterator seek request as incomplete, is disabled by default when max_skippable_internal_keys = 0. It is enabled only when max_skippable_internal_keys > 0.
This feature is based on the discussion mentioned in the PR https://github.com/facebook/rocksdb/pull/1084.
Closes https://github.com/facebook/rocksdb/pull/2000
Differential Revision: D4753223
Pulled By: sagar0
fbshipit-source-id: 1c973f7
Summary:
Previously, when DB write buffer size triggers, we always pick the CF with most data in its memtable to flush. This approach can minimize total flush happens. Change the behavior to always pick the oldest unflushed CF, which makes it the same behavior when max_total_wal_size hits. This approach will minimize size used by max_total_wal_size.
Closes https://github.com/facebook/rocksdb/pull/1987
Differential Revision: D4703214
Pulled By: siying
fbshipit-source-id: 9ff8b09
Summary:
PinnableSlice
Summary:
Currently the point lookup values are copied to a string provided by the
user. This incures an extra memcpy cost. This patch allows doing point lookup
via a PinnableSlice which pins the source memory location (instead of
copying their content) and releases them after the content is consumed
by the user. The old API of Get(string) is translated to the new API
underneath.
Here is the summary for improvements:
value 100 byte: 1.8% regular, 1.2% merge values
value 1k byte: 11.5% regular, 7.5% merge values
value 10k byte: 26% regular, 29.9% merge values
The improvement for merge could be more if we extend this approach to
pin the merge output and delay the full merge operation until the user
actually needs it. We have put that for future work.
PS:
Sometimes we observe a small decrease in performance when switching from
t5452014 to this patch but with the old Get(string) API. The d
Closes https://github.com/facebook/rocksdb/pull/1756
Differential Revision: D4391738
Pulled By: maysamyabandeh
fbshipit-source-id: 6f3edd3
Summary:
This PR is to support a way to iterate over all the keys that are just in memtables.
Closes https://github.com/facebook/rocksdb/pull/1953
Differential Revision: D4663500
Pulled By: sagar0
fbshipit-source-id: 144e177
Summary:
db_sst_test had been flaky occasionally in the following way: reached_max_space_on_compaction can in very rare cases be 0. This happens when the limit on maximum allowable space set using SetMaxAllowedSpaceUsage is hit during flush for all test db sizes (1,2,4,8 and 10MB).The fix clears the error returned when the the space limit is reached during flush. This ensures that the compaction call back will always be called. The runtime is increased slightly because the 1MB loop writes more data and hits the limit during multiple flushes until compaction is scheduled.
Closes https://github.com/facebook/rocksdb/pull/1861
Differential Revision: D4557396
Pulled By: lgalanis
fbshipit-source-id: ff778d1
Summary:
Relating to #1903:
In MaybeFlushColumnFamilies() we want to modify the 'getting_flushed' flag before releasing the db mutex when SwitchMemtable() is called.
The following 2 actions need to be atomic in MaybeFlushColumnFamilies()
- getting_flushed is false on oldest log
- we determine that all CFs can be flushed to successfully release oldest log
- we set getting_flushed = true on the oldest log.
-------
- getting_flushed is false on oldest log
- we determine that all CFs can NOT be flushed to successfully release oldest log
- we set unable_to_flush_oldest_log_ = true on the oldest log.
#### In the 2pc case:
T1 enters function but is unable to flush all CFs to release log
T1 sets unable_to_flush_oldest_log_ = true
T1 begins flushing all CFs possible
T2 enters function but is unable to flush all CFs to release log
T2 sees unable_to_flush_oldes_log_ has been set so exits
T3 enters function and will be able to flush all CFs to release oldest log
T3 sets getting_flushed = true on oldes
Closes https://github.com/facebook/rocksdb/pull/1909
Differential Revision: D4646235
Pulled By: reidHoruff
fbshipit-source-id: c8d0447
Summary:
This is a trivial fix for OOMs we've seen a few days ago in logdevice.
RocksDB get into the following state:
(1) Write throughput is too high for flushes to keep up. Compactions are out of the picture - automatic compactions are disabled, and for manual compactions we don't care that much if they fall behind. We write to many CFs, with only a few L0 sst files in each, so compactions are not needed most of the time.
(2) total_log_size_ is consistently greater than GetMaxTotalWalSize(). It doesn't get smaller since flushes are falling ever further behind.
(3) Total size of memtables is way above db_write_buffer_size and keeps growing. But the write_buffer_manager_->ShouldFlush() is not checked because (2) prevents it (for no good reason, afaict; this is what this commit fixes).
(4) Every call to WriteImpl() hits the MaybeFlushColumnFamilies() path. This keeps flushing the memtables one by one in order of increasing log file number.
(5) No write stalling trigger is hit. We rely on max_write_buffer_number
Closes https://github.com/facebook/rocksdb/pull/1893
Differential Revision: D4593590
Pulled By: yiwu-arbug
fbshipit-source-id: af79c5f
Summary:
The option has been deprecated for two years and has no effect. Removing.
Closes https://github.com/facebook/rocksdb/pull/1866
Differential Revision: D4555203
Pulled By: yiwu-arbug
fbshipit-source-id: c48f627
Summary:
Seems to me `has_unpersisted_data_` is read from read thread and write
from write thread concurrently without synchronization. Making it an
atomic.
I update the logic not because seeing any problem with it, but it just
feel confusing.
Closes https://github.com/facebook/rocksdb/pull/1869
Differential Revision: D4555837
Pulled By: yiwu-arbug
fbshipit-source-id: eff2ab8
Summary:
Remove disableDataSync, and another similarly named disable_data_sync options.
This is being done to simplify options, and also because the performance gains of this feature can be achieved by other methods.
Closes https://github.com/facebook/rocksdb/pull/1859
Differential Revision: D4541292
Pulled By: sagar0
fbshipit-source-id: 5b3a6ca
Summary:
Added method that returns approx num of entries as well as size for memtables.
Closes https://github.com/facebook/rocksdb/pull/1841
Differential Revision: D4511990
Pulled By: VitaliyLi
fbshipit-source-id: 9a4576e
Summary:
In theory, Get() can get a wrong result, if it races in a special with with flush. The bug can be reproduced in DBTest2.GetRaceFlush. Fix this bug by getting snapshot after referencing the super version.
Closes https://github.com/facebook/rocksdb/pull/1816
Differential Revision: D4475958
Pulled By: siying
fbshipit-source-id: bd9e67a
Summary:
merger.h was always a confusing name for me, simply give the file a better name
Closes https://github.com/facebook/rocksdb/pull/1836
Differential Revision: D4505357
Pulled By: IslamAbdelRahman
fbshipit-source-id: 07b28d8
Summary:
A current data race issue in Get() and Flush() can cause a Get() to return wrong results when a flush happened in the middle. Disable the test for now.
Closes https://github.com/facebook/rocksdb/pull/1813
Differential Revision: D4472310
Pulled By: siying
fbshipit-source-id: 5755ebd
Summary:
logs_.back() is called out of DB mutex, which can cause data race. We move the access into the DB mutex protection area.
Closes https://github.com/facebook/rocksdb/pull/1774
Reviewed By: AsyncDBConnMarkedDownDBException
Differential Revision: D4417472
Pulled By: AsyncDBConnMarkedDownDBException
fbshipit-source-id: 2da1f1e
Summary:
GetAndRefSuperVersion() should not be called again in the same thread before ReturnAndCleanupSuperVersion() is called.
If we have a compaction filter that is using DB::Get, This will happen
```
CompactFiles() {
GetAndRefSuperVersion() // -- first call
..
CompactionFilter() {
GetAndRefSuperVersion() // -- second call
ReturnAndCleanupSuperVersion()
}
..
ReturnAndCleanupSuperVersion()
}
```
We solve this issue in the same way Iterator is solving it, but using GetReferencedSuperVersion()
This was discovered in https://github.com/facebook/mysql-5.6/issues/427 by alxyang
Closes https://github.com/facebook/rocksdb/pull/1803
Differential Revision: D4460155
Pulled By: IslamAbdelRahman
fbshipit-source-id: 5e54322
