Commit Graph

5890 Commits

Author SHA1 Message Date
Maysam Yabandeh
54b434110e Builders for partition filter
Summary:
This is the second split of this pull request: https://github.com/facebook/rocksdb/pull/1891 which includes only the builder part. The testing will be included in the third split, where the reader is also included.
Closes https://github.com/facebook/rocksdb/pull/1952

Differential Revision: D4660272

Pulled By: maysamyabandeh

fbshipit-source-id: 36b3cf0
2017-03-07 13:54:12 -08:00
Sagar Vemuri
97edc72d39 Add a memtable-only iterator
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
2017-03-07 11:54:10 -08:00
Leonidas Galanis
72202962f9 fix db_sst_test flakiness
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
2017-03-07 11:24:13 -08:00
Matteo Merli
5f65dc8778 Expose DB::DeleteRange and WriteBath::DeleteRange in Java
Summary:
Added JNI wrapper from `DeleteRange` methods
Closes https://github.com/facebook/rocksdb/pull/1951

Differential Revision: D4657746

Pulled By: yiwu-arbug

fbshipit-source-id: 3fc7ab8
2017-03-06 22:24:26 -08:00
Reid Horuff
58b12dfe37 Set logs as getting flushed before releasing lock, race condition fix
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
2017-03-06 15:09:11 -08:00
Siying Dong
f8a4ea0206 Move db_test and external_sst_file_test out of Travis's MAC OS run
Summary:
After we have db_basic_test and external_sst_file_basic_test, we don't need to run db_test and external_sst_file_test in Travis's MAC OS run anymore. Move it out.
Closes https://github.com/facebook/rocksdb/pull/1940

Differential Revision: D4659361

Pulled By: siying

fbshipit-source-id: e64e291
2017-03-06 09:39:15 -08:00
Maysam Yabandeh
534581a356 Fix a bug in tests in options operator=
Summary:
Note: Using the default operator= is an unsafe approach for Options since it destructs shared_ptr in
the same order of their creation, in contrast to destructors which
destructs them in the opposite order of creation. One particular problme is
that the cache destructor might invoke callback functions that use Option
members such as statistics. To work around this problem, we manually call
destructor of table_facotry which eventually clears the block cache.
Closes https://github.com/facebook/rocksdb/pull/1950

Differential Revision: D4655473

Pulled By: maysamyabandeh

fbshipit-source-id: 6c4bbff
2017-03-05 18:09:09 -08:00
Maysam Yabandeh
a2f7a514d1 Refactoring
Summary:
This is the first split of https://github.com/facebook/rocksdb/pull/1891 and will be needed for the upcoming partitioned filter patch.
Closes https://github.com/facebook/rocksdb/pull/1949

Differential Revision: D4652152

Pulled By: maysamyabandeh

fbshipit-source-id: 9801778
2017-03-03 18:24:12 -08:00
Andrew Kryczka
2a5daa06f0 Add stderr log level for ldb backup commands
Summary:
Also extracted the common logic into a base class, BackupableCommand.
Closes https://github.com/facebook/rocksdb/pull/1939

Differential Revision: D4630121

Pulled By: ajkr

fbshipit-source-id: 04bb067
2017-03-03 13:24:15 -08:00
Andrew Kryczka
4561275c2d fix rate limiter test flakiness
Summary:
fix when elapsed time spans non-integral number of intervals since the rate limiter may still be drained during a partial interval.
Closes https://github.com/facebook/rocksdb/pull/1948

Differential Revision: D4651304

Pulled By: ajkr

fbshipit-source-id: b1f9e70
2017-03-03 11:09:11 -08:00
Andrew Kryczka
7c80a6d7d1 Statistic for how often rate limiter is drained
Summary:
This is the metric I plan to use for adaptive rate limiting. The statistics are updated only if the rate limiter is drained by flush or compaction. I believe (but am not certain) that this is the normal case.

The Statistics object is passed in RateLimiter::Request() to avoid requiring changes to client code, which would've been necessary if we passed it in the RateLimiter constructor.
Closes https://github.com/facebook/rocksdb/pull/1946

Differential Revision: D4646489

Pulled By: ajkr

fbshipit-source-id: d8e0161
2017-03-02 17:54:15 -08:00
Andrew Kryczka
0ad5af42d0 Clarify VerifyBackup behavior
Summary:
It's non-obvious to users that using the same backup engine for creating/verifying provides better results than using separate backup engines, so add a comment in header.
Closes https://github.com/facebook/rocksdb/pull/1942

Differential Revision: D4637865

Pulled By: ajkr

fbshipit-source-id: e6efe24
2017-03-02 17:24:11 -08:00
Aaron Gao
6fb9013441 sanitize readahead when direct read enabled
Summary:
no readahead:
readseq      :       8.438 micros/op 118510 ops/sec;   13.1 MB/s
sanitize to 10MB:
readseq      :       6.051 micros/op 165248 ops/sec;   18.3 MB/s
Closes https://github.com/facebook/rocksdb/pull/1945

Differential Revision: D4645811

Pulled By: lightmark

fbshipit-source-id: 5d63770
2017-03-02 17:24:11 -08:00
Islam AbdelRahman
f89b3893c0 Remove skip_table_builder_flush and default it to true
Summary:
This option is needed to be enabled for Direct IO
and I cannot think of a reason where we need to disable it

remove it and default it to true
Closes https://github.com/facebook/rocksdb/pull/1944

Differential Revision: D4641088

Pulled By: IslamAbdelRahman

fbshipit-source-id: d7085b9
2017-03-02 16:54:10 -08:00
Andrew Kryczka
cc253982dd Use more default options in db_bench
Summary:
The default behavior was too weird because, previously, we got the L0 file size limit (64MB) from Options default and L1+ file size limit (2MB) from the hardcoded value. We should get both from Options default.
Closes https://github.com/facebook/rocksdb/pull/1943

Differential Revision: D4640301

Pulled By: ajkr

fbshipit-source-id: fd8c0fd
2017-03-02 10:54:11 -08:00
Siying Dong
8432bcf555 Make compaction_pri settable through option string
Summary: Closes https://github.com/facebook/rocksdb/pull/1941

Differential Revision: D4637253

Pulled By: siying

fbshipit-source-id: a59dcdb
2017-03-02 10:24:12 -08:00
Siying Dong
d5b607a43f Make db_wal_test slightly faster
Summary:
Avoid to run db_wal_test in all the DB test options, and some small changes.
Closes https://github.com/facebook/rocksdb/pull/1921

Differential Revision: D4622054

Pulled By: siying

fbshipit-source-id: 890fd64
2017-02-28 17:39:10 -08:00
Siying Dong
ba4c77bd6b Divide external_sst_file_test
Summary:
Separate the platform dependent tests from external_sst_file_test. Only those tests need to run on platforms like OSX
Closes https://github.com/facebook/rocksdb/pull/1923

Differential Revision: D4622461

Pulled By: siying

fbshipit-source-id: d2d6f04
2017-02-28 14:24:11 -08:00
Aaron Gao
e877afa08b Remove bulk loading and auto_roll_logger in rocksdb_lite
Summary:
shrink lite size
Closes https://github.com/facebook/rocksdb/pull/1929

Differential Revision: D4622059

Pulled By: siying

fbshipit-source-id: 050b796
2017-02-28 11:09:11 -08:00
xiusir
90d8355075 Fix the wrong address for PREFETCH in DynamicBloom::Prefetch
Summary:
- Change data_[b] to data_[b / 8] in DynamicBloom::Prefetch, as b means the b-th bit in data_ and data_[b / 8] is the proper byte in data_.
Closes https://github.com/facebook/rocksdb/pull/1935

Differential Revision: D4628696

Pulled By: siying

fbshipit-source-id: bc5a0c6
2017-02-28 10:39:11 -08:00
Islam AbdelRahman
08864df212 Move advanced column family options to advanced_options.h
Summary:
For the sake of making our options simpler, we should keep options.h as simple as possible and move more advanced/less common options to advaned_options.h

I started with ColumnFamilyOptions and also did some re-ordering

I have moved all ColumnFamilyOptions to advanced_options.h and only left these options in options.h

```
const Comparator* comparator = BytewiseComparator();
std::shared_ptr<MergeOperator> merge_operator = nullptr;
const CompactionFilter* compaction_filter = nullptr;
std::shared_ptr<CompactionFilterFactory> compaction_filter_factory = nullptr;
size_t write_buffer_size = 64 << 20;
CompressionType compression;
int level0_file_num_compaction_trigger = 4;
bool disable_auto_compactions = false;
```
Please feel free to comment on specific options if you think they should be advanced or should not be
Closes https://github.com/facebook/rocksdb/pull/1847

Differential Revision: D4519996

Pulled By: IslamAbdelRahman

fbshipit-source-id: abebd9a
2017-02-27 17:54:14 -08:00
Peter (Stig) Edwards
2ca2059f66 Get unique_ptr to use delete[] for char[] in DumpMallocStats
Summary:
Avoid mismatched free() / delete / delete [] in DumpMallocStats
Closes https://github.com/facebook/rocksdb/pull/1927

Differential Revision: D4622045

Pulled By: siying

fbshipit-source-id: 1131b30
2017-02-27 17:39:12 -08:00
Tamir Duberstein
253799c06d Add missing include for abort()
Summary:
Fixes #1233 (again).
Closes https://github.com/facebook/rocksdb/pull/1931

Differential Revision: D4625289

Pulled By: ajkr

fbshipit-source-id: 70e774e
2017-02-27 17:24:13 -08:00
Adam Retter
c6d464a9da Fixed various memory leaks and Java 8 JNI Compatibility
Summary:
I have manually audited the entire RocksJava code base.

Sorry for the large pull-request, I have broken it down into many small atomic commits though.

My initial intention was to fix the warnings that appear when running RocksJava on Java 8 with `-Xcheck:jni`, for example when running `make jtest` you would see many errors similar to:

```
WARNING in native method: JNI call made without checking exceptions when required to from CallObjectMethod
WARNING in native method: JNI call made without checking exceptions when required to from CallVoidMethod
WARNING in native method: JNI call made without checking exceptions when required to from CallStaticVoidMethod
...
```

A few of those warnings still remain, however they seem to come directly from the JVM and are not directly related to RocksJava; I am in contact with the OpenJDK hostpot-dev mailing list about these - http://mail.openjdk.java.net/pipermail/hotspot-dev/2017-February/025981.html.

As a result of fixing these, I realised we were not r
Closes https://github.com/facebook/rocksdb/pull/1890

Differential Revision: D4591758

Pulled By: siying

fbshipit-source-id: 7f7fdf4
2017-02-27 16:39:12 -08:00
Islam AbdelRahman
be3e5568be Fix unaligned reads in read cache
Summary:
- Fix unaligned reads in read cache by using RandomAccessFileReader
- Allow read cache flags in db_bench
Closes https://github.com/facebook/rocksdb/pull/1916

Differential Revision: D4610885

Pulled By: IslamAbdelRahman

fbshipit-source-id: 2aa1dc8
2017-02-27 13:09:12 -08:00
Siying Dong
8ad0fcdf99 Separate small subset tests in DBTest
Summary:
Separate a smal subset of tests in DBTest to DBBasicTest. Tests in DBTest don't have to run in CI tests on platforms like OSX, as long as they are covered by Linux.
Closes https://github.com/facebook/rocksdb/pull/1924

Differential Revision: D4616702

Pulled By: siying

fbshipit-source-id: 13e6549
2017-02-27 12:24:11 -08:00
Siying Dong
6c951c43c7 Run fewer tests in OSX
Summary:
Travis is short of OSX resource. Try to move platform independent test suites out of OSX
Closes https://github.com/facebook/rocksdb/pull/1922

Differential Revision: D4616070

Pulled By: siying

fbshipit-source-id: 786342c
2017-02-27 11:09:20 -08:00
Aaron Gao
f7997f1341 add direct I/O to version notes 5.2.0
Summary:
let users know this feature is ready
Closes https://github.com/facebook/rocksdb/pull/1915

Differential Revision: D4610842

Pulled By: lightmark

fbshipit-source-id: d102772
2017-02-23 22:09:12 -08:00
Siying Dong
3b8ba703cb Fix flaky DBTestUniversalCompaction.UniversalCompactionTrivialMoveTest2
Summary:
A previous fix to DBTestUniversalCompaction.UniversalCompactionTrivialMoveTest2 didn't address the right problem. The problem is L0->L0 compaction is not trivial move in the scenario, not parallel compactions. Fix this.
Closes https://github.com/facebook/rocksdb/pull/1911

Differential Revision: D4608955

Pulled By: siying

fbshipit-source-id: 7a712cb
2017-02-23 18:39:13 -08:00
Siying Dong
96c7e1504a Fix Java build
Summary:
The PATH update should put the Java path in the beginning, rather than the end. Otherwise, it will be overwritten. Also upgrade the Java version.
Closes https://github.com/facebook/rocksdb/pull/1912

Differential Revision: D4609854

Pulled By: siying

fbshipit-source-id: 3dc04f2
2017-02-23 18:39:13 -08:00
Siying Dong
e0b87afc70 Black list some slow valgrind tests
Summary:
valgrind tests always timeout with parallel run. Black list some slowest ones. It is better to run fewer tests than always have the tests timeout.
Closes https://github.com/facebook/rocksdb/pull/1908

Differential Revision: D4607875

Pulled By: siying

fbshipit-source-id: 7062664
2017-02-23 16:09:11 -08:00
Andrew Kryczka
e67232cff5 Handle failed Finish() in SST file writer
Summary:
The assertion in Abandon() fails when called after Finish() fails. Finish() already closes the builder so there's no need to call Abandon().
Closes https://github.com/facebook/rocksdb/pull/1901

Differential Revision: D4601373

Pulled By: ajkr

fbshipit-source-id: e5678be
2017-02-23 15:39:16 -08:00
Siying Dong
8efb5ffa2a [rocksdb][PR] Remove option min_partial_merge_operands and verify_checksums_in_comp…
Summary:
…action

 The two options, min_partial_merge_operands and verify_checksums_in_compaction, are not seldom used. Remove them to reduce the total number of options. Also remove them from Java and C interface.
Closes https://github.com/facebook/rocksdb/pull/1902

Differential Revision: D4601219

Pulled By: siying

fbshipit-source-id: aad4cb2
2017-02-23 15:09:12 -08:00
Siying Dong
1ba2804b7f Remove XFunc tests
Summary:
Xfunc is hardly used. Remove it to keep the code simple.
Closes https://github.com/facebook/rocksdb/pull/1905

Differential Revision: D4603220

Pulled By: siying

fbshipit-source-id: 731f96d
2017-02-23 12:09:11 -08:00
Aaron Gao
e7d902e693 add direct_io and compaction_readahead_size in db_stress
Summary:
add direct_io and compaction_readahead_size in db_stress
test direct_io under db_stress with compaction_readahead_size enabled to capture bugs found in production.
`./db_stress --allow_concurrent_memtable_write=0 --use_direct_reads --use_direct_writes --compaction_readahead_size=4096`
Closes https://github.com/facebook/rocksdb/pull/1906

Differential Revision: D4604514

Pulled By: IslamAbdelRahman

fbshipit-source-id: ebbf0ee
2017-02-23 11:39:14 -08:00
Aaron Gao
1ef5f50e84 detect logical sector size
Summary:
querying logical sector size from the device instead of hardcoding it for linux platform.
Closes https://github.com/facebook/rocksdb/pull/1875

Differential Revision: D4591946

Pulled By: ajkr

fbshipit-source-id: 4e9805c
2017-02-23 11:25:36 -08:00
Andrew Kryczka
ed50308d20 check backup directory exists before listing children
Summary:
InsertPathnameToSizeBytes() is called on shared/ and shared_checksum/ directories, which only exist for certain configurations. If we try to list a non-existent directory's contents, some Envs will dump an error message. Let's avoid this by checking whether the directory exists before listing its contents.
Closes https://github.com/facebook/rocksdb/pull/1895

Differential Revision: D4596301

Pulled By: ajkr

fbshipit-source-id: c809679
2017-02-23 10:54:10 -08:00
Giuseppe Ottaviano
4d7c06cedf Make WriteBatchWithIndex moveble
Summary:
`WriteBatchWithIndex` has an incorrect implicitly-generated move constructor (it will copy the pointer causing a double-free on destruction). Just switch to `unique_ptr` so we get correct move semantics for free.
Closes https://github.com/facebook/rocksdb/pull/1899

Differential Revision: D4598896

Pulled By: ajkr

fbshipit-source-id: 2373d47
2017-02-22 17:54:11 -08:00
Andrew Kryczka
5040414e6f Gracefully handle previous backup interrupted
Summary:
As the last step in backup creation, the .tmp directory is renamed omitting the .tmp suffix. In case the process terminates before this, the .tmp directory will be left behind. Even if this happens, we want future backups to succeed, so I added some checks/cleanup for this case.
Closes https://github.com/facebook/rocksdb/pull/1896

Differential Revision: D4597323

Pulled By: ajkr

fbshipit-source-id: 48900d8
2017-02-22 17:39:12 -08:00
Aaron Gao
f206af56fc add use_direct_io() to ReadaheadRandomAccessFile
Summary:
Missing this function will cause RandomAccessFileReader not doing alignment in Direct IO mode, which introduce an IOError: invalid argument.
Closes https://github.com/facebook/rocksdb/pull/1900

Differential Revision: D4601261

Pulled By: lightmark

fbshipit-source-id: c3eadf1
2017-02-22 14:54:11 -08:00
Aaron Gao
0824934423 truncate patch
Summary:
omit the override for the previous commit
Closes https://github.com/facebook/rocksdb/pull/1898

Differential Revision: D4598743

Pulled By: lightmark

fbshipit-source-id: f98a378
2017-02-22 10:39:11 -08:00
Aaron Gao
286a36db7f posix writablefile truncate
Summary:
we occasionally missing this call so the file size will be wrong
Closes https://github.com/facebook/rocksdb/pull/1894

Differential Revision: D4598446

Pulled By: lightmark

fbshipit-source-id: 42b6ef5
2017-02-22 10:09:14 -08:00
Daniel Black
f0879e4c39 Page size isn't always 4k on linux
Summary:
Some places autodetected. These are the two places that didn't.

closes #1498

Still unsure if the following instances of 4 * 1024 need fixing in:
util/io_posix.h
include/rocksdb/table.h (appears to be blocksize and different)
utilities/persistent_cache/block_cache_tier.cc
utilities/persistent_cache/persistent_cache_test.h
include/rocksdb/env.h
util/env_posix.cc
db/column_family.cc
Closes https://github.com/facebook/rocksdb/pull/1499

Differential Revision: D4593640

Pulled By: yiwu-arbug

fbshipit-source-id: efc48de
2017-02-21 16:39:14 -08:00
Mike Kolupaev
18eeb7b90e Fix interference between max_total_wal_size and db_write_buffer_size checks
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
2017-02-21 16:09:10 -08:00
Islam AbdelRahman
1560b2f5f0 Temporarly return deprecated functions to fix MongoRocks build
Summary:
MongoRocks is still using some deprecated functions, return them temporarily
Closes https://github.com/facebook/rocksdb/pull/1892

Differential Revision: D4592451

Pulled By: IslamAbdelRahman

fbshipit-source-id: 5e6be3e
2017-02-21 12:54:11 -08:00
Aaron Gao
2a0f3d0de1 level compaction expansion
Summary:
reimplement the compaction expansion on lower level.

Considering such a case:
input level file: 1[B E] 2[F G] 3[H I] 4 [J M]
output level file: 5[A C] 6[D K] 7[L O]

If we initially pick file 2, now we will compact file 2 and 6. But we can safely compact 2, 3 and 6 without expanding the output level.

The previous code is messy and wrong.

In this diff, I first determine the input range [a, b], and output range [c, d],
then we get the range [e,f] = [min(a, c), max(b, d] and put all eligible clean-cut files within [e, f] into this compaction.

**Note: clean-cut means the files don't have the same user key on the boundaries of some files that are not chosen in this compaction**.
Closes https://github.com/facebook/rocksdb/pull/1760

Differential Revision: D4395564

Pulled By: lightmark

fbshipit-source-id: 2dc2c5c
2017-02-21 10:24:17 -08:00
Yulia Kartseva
ebc8a79980 alignment is on in ReadaheadRandomAccessFile::Read()
Summary: Closes https://github.com/facebook/rocksdb/pull/1857

Differential Revision: D4534518

Pulled By: wat-ze-hex

fbshipit-source-id: b456946
2017-02-18 12:09:12 -08:00
Yi Wu
381fd32247 Remove timeout_hint_us from WriteOptions
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
2017-02-17 15:24:17 -08:00
Islam AbdelRahman
fce7a6e196 Fail IngestExternalFile when bg_error_ exists
Summary:
Fail IngestExternalFile() when bg_error_ exists
Closes https://github.com/facebook/rocksdb/pull/1881

Differential Revision: D4580621

Pulled By: IslamAbdelRahman

fbshipit-source-id: 1194913
2017-02-17 13:39:17 -08:00
Marcin Dlugajczyk
a618a16f44 New subcode for IOError to detect the ESTALE errno
Summary:
I'd like to propose a patch to expose a new IOError type with subcode kStaleFile to allow to detect when ESTALE error is returned. This allows the rocksdb consumers to handle this error separately from other IOErrors.

I've also added a missing string representation for the kDeadlock subcode, I believe calling ToString() on Status object with that subcode would result in an out of band access in the msgs array,

Please let me know if you have any questions or would like me to make any changes to this pull request.
Closes https://github.com/facebook/rocksdb/pull/1748

Differential Revision: D4387675

Pulled By: IslamAbdelRahman

fbshipit-source-id: 67feb13
2017-02-17 10:54:13 -08:00