Summary:
This is a followup to https://github.com/facebook/rocksdb/issues/6646. In that PR, for simplicity I just appended a comparison against the 0th restart key in case `BinarySeek()`'s binary search landed at index 0. As a result there were `2/(N+1) + log_2(N)` key comparisons. This PR does it differently. Now we expand the binary search range by one so it also covers the case where target is at or before the restart key at index 0. As a result, it involves `log_2(N+1)` key comparisons.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7068
Test Plan:
ran readrandom with mostly default settings and counted key comparisons
using `PerfContext`.
before: `user_key_comparison_count = 28881965`
after: `user_key_comparison_count = 27823245`
setup command:
```
$ TEST_TMPDIR=/dev/shm/dbbench ./db_bench -benchmarks=fillrandom,compact -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -max_background_jobs=12 -level_compaction_dynamic_level_bytes=true -num=10000000
```
benchmark command:
```
$ TEST_TMPDIR=/dev/shm/dbbench/ ./db_bench -use_existing_db=true -benchmarks=readrandom -disable_auto_compactions=true -num=10000000 -compression_type=none -reads=1000000 -perf_level=3
```
Reviewed By: anand1976
Differential Revision: D22357032
Pulled By: ajkr
fbshipit-source-id: 8b01e9c1c2a4e9d02fc9dfe16c1cc0327f8bdf24
Summary:
When format_version is high enough to support user-key and
there are index entries for same user key that spans multiple data
blocks then it changes from user-key mode to internal-key mode. But the
flush policy is not reset to point to Block Builder of internal-keys.
After this switch, no entries are added to user key index partition
result, thus it never triggers flushing the block.
Fix: 1. After adding the entry in sub_builder_index_, if there is a switch
from user-key to internal-key, then flush policy is updated to point to
Block Builder of internal-keys index partition.
2. Set sub_builder_index_->seperator_is_key_plus_seq_ = true if
seperator_is_key_plus_seq_ is set to true so that subsequent partitions
can also use internal key mode.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7096
Test Plan: make check -j64
Reviewed By: ajkr
Differential Revision: D22416598
Pulled By: akankshamahajan15
fbshipit-source-id: 01fc2dc07ea1b32f8fb803995ebe6e9a3fbe67ac
Summary:
Currently, `EventListener` in listner.h only have callback functions for file read and write. One may favor extended callback functions for more file I/O operations like flush, sync and close. This PR tries to add those interface and have them called when appropriate throughout the code base.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7055
Test Plan:
Write an experimental listener with those new callback functions with log output in them; run experiments and check logs to see those functions are actually called.
Default test suits `make check` should also be included.
Reviewed By: riversand963
Differential Revision: D22380624
Pulled By: roghnin
fbshipit-source-id: 4121491d45c2c2aae8c255e7998090559a241c6a
Summary:
When table file checksums are enabled and stored in the DB manifest by using the RocksDB default crc32c checksum function, BackupEngine will calculate the crc32c checksum of the file to be copied and compare the calculated result with the one stored in the DB manifest before copying the file to the backup directory.
After copying to the backup directory, BackupEngine will verify the checksum of the copied file with the one calculated before copying. This helps detect some rare corruption events such as bit-flips during the copying process.
No verification with checksums in DB manifest will be performed if the table file checksum function is not the RocksDB default crc32c checksum function.
In addition, If `share_table_files` and `share_files_with_checksum` are true, BackupEngine will compare the checksums computed before and after copying of the table files.
Corresponding tests are added.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7015
Test Plan: Passed make check
Reviewed By: pdillinger
Differential Revision: D22165732
Pulled By: gg814
fbshipit-source-id: ee0e8cc397c455eba64545c29380b9d9853588ec
Summary:
When format_version is high enough to support user-key and there are index entries for same user key that spans multiple data blocks then it changes from user-key mode to internal-key mode. But the flush policy is not reset to point to Block Builder of internal-keys. After this switch, no entries are added to user key index partition result, thus it never triggers flushing the block.
Fix: After adding the entry in sub_builder_index_, if there is a switch from user-key to internal-key, then flush policy is updated to point to Block Builder of internal-keys index partition.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7022
Test Plan:
1. make check -j64
2. Added one unit test case
Reviewed By: ajkr
Differential Revision: D22197734
Pulled By: akankshamahajan15
fbshipit-source-id: d87e9e46bccab8e896ee6979d6b79c51f73d479e
Summary:
`bool BackupableDBOptions::new_naming_for_backup_files` is updated to `BackupTableNameOption BackupableDBOptions::share_files_with_checksum_naming`, where `BackupTableNameOption` is an `enum` type with two enumerators `kChecksumAndFileSize` and `kChecksumAndFileSize`. This opens up possibilities of extenting the current naming scheme for backup table files. By default, `BackupTableNameOption BackupableDBOptions::share_files_with_checksum_naming` is set to `kChecksumAndDbSessionId`.
Revert `BackupEngine::VerifyBackup` to only check file sizes by default.
Also fix the construction of the `SstFileDumper` in `GetFileDbIdentities` by setting a proper `Env` of the `Options` passed in the constructor.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7032
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D22237763
Pulled By: gg814
fbshipit-source-id: 466902a4e731babd64e30f0e82ca1aa82962e52e
Summary:
Added compaction filter support for BlobDB non-TTL values. Same as vanilla RocksDB, user compaction filter applies to all k/v pairs of the compaction for non-TTL values. It honors `min_blob_size`, which potentially results value transitions between inlined data and stored-in-blob data when size of value is changed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6850
Reviewed By: siying
Differential Revision: D22263487
Pulled By: ltamasi
fbshipit-source-id: 8fc03f8cde2a5c831e63b436b3dbf1b7f90939e8
Summary:
A parameter `verify_with_checksum` is added to `BackupEngine::VerifyBackup`, which is true by default. So now `BackupEngine::VerifyBackup` verifies backup files with checksum AND file size by default. When `verify_with_checksum` is false, `BackupEngine::VerifyBackup` only compares file sizes to verify backup files.
Also add a test for the case when corruption does not change the file size.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7014
Test Plan: Passed backupable_db_test
Reviewed By: zhichao-cao
Differential Revision: D22165590
Pulled By: gg814
fbshipit-source-id: 606a7450714e868bceb38598c89fd356c6004f4f
Summary:
`DB::OpenForReadOnly()` now returns `Status::NotFound` when the specified DB directory does not exist. Previously the error returned depended on the underlying `Env`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7023
Reviewed By: ajkr
Differential Revision: D22207845
Pulled By: gg814
fbshipit-source-id: f35830811a0e67efb0ee82eda3a9739bc526baba
Summary:
`BackupableDBOptions::new_naming_for_backup_files` is added. This option is false by default. When it is true, backup table filenames under directory shared_checksum are of the form `<file_number>_<crc32c>_<db_session_id>.sst`.
Note that when this option is true, it comes into effect only when both `share_files_with_checksum` and `share_table_files` are true.
Three new test cases are added.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6997
Test Plan: Passed make check.
Reviewed By: ajkr
Differential Revision: D22098895
Pulled By: gg814
fbshipit-source-id: a1d9145e7fe562d71cde7ac995e17cb24fd42e76
Summary:
This PR provides preliminary support for handling IO error during MANIFEST write.
File write/sync is not guaranteed to be atomic. If we encounter an IOError while writing/syncing to the MANIFEST file, we cannot be sure about the state of the MANIFEST file. The version edits may or may not have reached the file. During cleanup, if we delete the newly-generated SST files referenced by the pending version edit(s), but the version edit(s) actually are persistent in the MANIFEST, then next recovery attempt will process the version edits(s) and then fail since the SST files have already been deleted.
One approach is to truncate the MANIFEST after write/sync error, so that it is safe to delete the SST files. However, file truncation may not be supported on certain file systems. Therefore, we take the following approach.
If an IOError is detected during MANIFEST write/sync, we disable file deletions for the faulty database. Depending on whether the IOError is retryable (set by underlying file system), either RocksDB or application can call `DB::Resume()`, or simply shutdown and restart. During `Resume()`, RocksDB will try to switch to a new MANIFEST and write all existing in-memory version storage in the new file. If this succeeds, then RocksDB may proceed. If all recovery is completed, then file deletions will be re-enabled.
Note that multiple threads can call `LogAndApply()` at the same time, though only one of them will be going through the process MANIFEST write, possibly batching the version edits of other threads. When the leading MANIFEST writer finishes, all of the MANIFEST writing threads in this batch will have the same IOError. They will all call `ErrorHandler::SetBGError()` in which file deletion will be disabled.
Possible future directions:
- Add an `ErrorContext` structure so that it is easier to pass more info to `ErrorHandler`. Currently, as in this example, a new `BackgroundErrorReason` has to be added.
Test plan (dev server):
make check
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6949
Reviewed By: anand1976
Differential Revision: D22026020
Pulled By: riversand963
fbshipit-source-id: f3c68a2ef45d9b505d0d625c7c5e0c88495b91c8
Summary:
New experimental option BBTO::optimize_filters_for_memory builds
filters that maximize their use of "usable size" from malloc_usable_size,
which is also used to compute block cache charges.
Rather than always "rounding up," we track state in the
BloomFilterPolicy object to mix essentially "rounding down" and
"rounding up" so that the average FP rate of all generated filters is
the same as without the option. (YMMV as heavily accessed filters might
be unluckily lower accuracy.)
Thus, the option near-minimizes what the block cache considers as
"memory used" for a given target Bloom filter false positive rate and
Bloom filter implementation. There are no forward or backward
compatibility issues with this change, though it only works on the
format_version=5 Bloom filter.
With Jemalloc, we see about 10% reduction in memory footprint (and block
cache charge) for Bloom filters, but 1-2% increase in storage footprint,
due to encoding efficiency losses (FP rate is non-linear with bits/key).
Why not weighted random round up/down rather than state tracking? By
only requiring malloc_usable_size, we don't actually know what the next
larger and next smaller usable sizes for the allocator are. We pick a
requested size, accept and use whatever usable size it has, and use the
difference to inform our next choice. This allows us to narrow in on the
right balance without tracking/predicting usable sizes.
Why not weight history of generated filter false positive rates by
number of keys? This could lead to excess skew in small filters after
generating a large filter.
Results from filter_bench with jemalloc (irrelevant details omitted):
(normal keys/filter, but high variance)
$ ./filter_bench -quick -impl=2 -average_keys_per_filter=30000 -vary_key_count_ratio=0.9
Build avg ns/key: 29.6278
Number of filters: 5516
Total size (MB): 200.046
Reported total allocated memory (MB): 220.597
Reported internal fragmentation: 10.2732%
Bits/key stored: 10.0097
Average FP rate %: 0.965228
$ ./filter_bench -quick -impl=2 -average_keys_per_filter=30000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory
Build avg ns/key: 30.5104
Number of filters: 5464
Total size (MB): 200.015
Reported total allocated memory (MB): 200.322
Reported internal fragmentation: 0.153709%
Bits/key stored: 10.1011
Average FP rate %: 0.966313
(very few keys / filter, optimization not as effective due to ~59 byte
internal fragmentation in blocked Bloom filter representation)
$ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000 -vary_key_count_ratio=0.9
Build avg ns/key: 29.5649
Number of filters: 162950
Total size (MB): 200.001
Reported total allocated memory (MB): 224.624
Reported internal fragmentation: 12.3117%
Bits/key stored: 10.2951
Average FP rate %: 0.821534
$ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory
Build avg ns/key: 31.8057
Number of filters: 159849
Total size (MB): 200
Reported total allocated memory (MB): 208.846
Reported internal fragmentation: 4.42297%
Bits/key stored: 10.4948
Average FP rate %: 0.811006
(high keys/filter)
$ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000000 -vary_key_count_ratio=0.9
Build avg ns/key: 29.7017
Number of filters: 164
Total size (MB): 200.352
Reported total allocated memory (MB): 221.5
Reported internal fragmentation: 10.5552%
Bits/key stored: 10.0003
Average FP rate %: 0.969358
$ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory
Build avg ns/key: 30.7131
Number of filters: 160
Total size (MB): 200.928
Reported total allocated memory (MB): 200.938
Reported internal fragmentation: 0.00448054%
Bits/key stored: 10.1852
Average FP rate %: 0.963387
And from db_bench (block cache) with jemalloc:
$ ./db_bench -db=/dev/shm/dbbench.no_optimize -benchmarks=fillrandom -format_version=5 -value_size=90 -bloom_bits=10 -num=2000000 -threads=8 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false
$ ./db_bench -db=/dev/shm/dbbench -benchmarks=fillrandom -format_version=5 -value_size=90 -bloom_bits=10 -num=2000000 -threads=8 -optimize_filters_for_memory -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false
$ (for FILE in /dev/shm/dbbench.no_optimize/*.sst; do ./sst_dump --file=$FILE --show_properties | grep 'filter block' ; done) | awk '{ t += $4; } END { print t; }'
17063835
$ (for FILE in /dev/shm/dbbench/*.sst; do ./sst_dump --file=$FILE --show_properties | grep 'filter block' ; done) | awk '{ t += $4; } END { print t; }'
17430747
$ #^ 2.1% additional filter storage
$ ./db_bench -db=/dev/shm/dbbench.no_optimize -use_existing_db -benchmarks=readrandom,stats -statistics -bloom_bits=10 -num=2000000 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false -duration=10 -cache_index_and_filter_blocks -cache_size=1000000000
rocksdb.block.cache.index.add COUNT : 33
rocksdb.block.cache.index.bytes.insert COUNT : 8440400
rocksdb.block.cache.filter.add COUNT : 33
rocksdb.block.cache.filter.bytes.insert COUNT : 21087528
rocksdb.bloom.filter.useful COUNT : 4963889
rocksdb.bloom.filter.full.positive COUNT : 1214081
rocksdb.bloom.filter.full.true.positive COUNT : 1161999
$ #^ 1.04 % observed FP rate
$ ./db_bench -db=/dev/shm/dbbench -use_existing_db -benchmarks=readrandom,stats -statistics -bloom_bits=10 -num=2000000 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false -optimize_filters_for_memory -duration=10 -cache_index_and_filter_blocks -cache_size=1000000000
rocksdb.block.cache.index.add COUNT : 33
rocksdb.block.cache.index.bytes.insert COUNT : 8448592
rocksdb.block.cache.filter.add COUNT : 33
rocksdb.block.cache.filter.bytes.insert COUNT : 18220328
rocksdb.bloom.filter.useful COUNT : 5360933
rocksdb.bloom.filter.full.positive COUNT : 1321315
rocksdb.bloom.filter.full.true.positive COUNT : 1262999
$ #^ 1.08 % observed FP rate, 13.6% less memory usage for filters
(Due to specific key density, this example tends to generate filters that are "worse than average" for internal fragmentation. "Better than average" cases can show little or no improvement.)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6427
Test Plan: unit test added, 'make check' with gcc, clang and valgrind
Reviewed By: siying
Differential Revision: D22124374
Pulled By: pdillinger
fbshipit-source-id: f3e3aa152f9043ddf4fae25799e76341d0d8714e
Summary:
EncryptEnv class is both declared and defined within env_encryption.cc. This makes it really tough to derive new classes from that base.
This branch moves declaration of the class to rocksdb/env_encryption.h. The change facilitates making new encryption modules (such as an upcoming openssl AES CTR pull request) possible / easy.
The only coding change was to add the EncryptEnv object to env_basic_test.cc.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6830
Reviewed By: riversand963
Differential Revision: D21706593
Pulled By: ajkr
fbshipit-source-id: 64d2da95a1569ceeb9b1549c3bec5404cf4c89f0
Summary:
The bug fixed in https://github.com/facebook/rocksdb/pull/1816/ is now applicable to iterator too. This was not an issue but https://github.com/facebook/rocksdb/pull/2886 caused the regression. If a put and DB flush happens just between iterator to get latest sequence number and getting super version, empty result for the key or an older value can be returned, which is wrong.
Fix it in the same way as the fix in https://github.com/facebook/rocksdb/issues/1816, that is to get the sequence number after referencing the super version.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6973
Test Plan: Will run stress tests for a while to make sure there is no general regression.
Reviewed By: ajkr
Differential Revision: D22029348
fbshipit-source-id: 94390f93630906796d6e2fec321f44a920953fd1
Summary:
https://github.com/facebook/rocksdb/issues/5411 refactored `VersionSet::Recover` but introduced a bug, explained as follows.
Before, once a checksum mismatch happens, `reporter` will set `s` to be non-ok. Therefore, Recover will stop processing the MANIFEST any further.
```
// Correct
// Inside Recover
LogReporter reporter;
reporter.status = &s;
log::Reader reader(..., reporter);
while (reader.ReadRecord() && s.ok()) {
...
}
```
The bug is that, the local variable `s` in `ReadAndRecover` won't be updated by `reporter` while reading the MANIFEST. It is possible that the reader sees a checksum mismatch in a record, but `ReadRecord` retries internally read and finds the next valid record. The mismatched record will be ignored and no error is reported.
```
// Incorrect
// Inside Recover
LogReporter reporter;
reporter.status = &s;
log::Reader reader(..., reporter);
s = ReadAndRecover(reader, ...);
// Inside ReadAndRecover
Status s; // Shadows the s in Recover.
while (reader.ReadRecord() && s.ok()) {
...
}
```
`LogReporter` can use a separate `log_read_status` to track the errors while reading the MANIFEST. RocksDB can process more MANIFEST entries only if `log_read_status.ok()`.
Test plan (devserver):
make check
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6996
Reviewed By: ajkr
Differential Revision: D22105746
Pulled By: riversand963
fbshipit-source-id: b22f717a423457a41ca152a242abbb64cf91fc38
Summary:
Compressed block cache is disabled in https://github.com/facebook/rocksdb/pull/4650 for no good reason. Re-enable it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6990
Test Plan: Add a unit test to make sure a general function works with read-only DB + compressed block cache.
Reviewed By: ltamasi
Differential Revision: D22072755
fbshipit-source-id: 2a55df6363de23a78979cf6c747526359e5dc7a1
Summary:
`db_id` and `db_session_id` are now part of the table properties for all formats and stored in SST files. This adds about 99 bytes to each new SST file.
The `TablePropertiesNames` for these two identifiers are `rocksdb.creating.db.identity` and `rocksdb.creating.session.identity`.
In addition, SST files generated from SstFileWriter and Repairer have DB identity “SST Writer” and “DB Repairer”, respectively. Their DB session IDs are generated in the same way as `DB::GetDbSessionId`.
A table property test is added.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6983
Test Plan: make check and some manual tests.
Reviewed By: zhichao-cao
Differential Revision: D22048826
Pulled By: gg814
fbshipit-source-id: afdf8c11424a6f509b5c0b06dafad584a80103c9
Summary:
In best-efforts recovery, an error that is not Corruption or IOError::kNotFound or IOError::kPathNotFound will be overwritten silently. Fix this by checking all non-ok cases and return early.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6989
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D22071418
Pulled By: riversand963
fbshipit-source-id: 5a4ea5dfb1a41f41c7a3fdaf62b163007b42f04b
Summary:
Best-efforts recovery does not check the content of CURRENT file to determine which MANIFEST to recover from. However, it still checks the presence of CURRENT file to determine whether to create a new DB during `open()`. Therefore, we can tweak the logic in `open()` a little bit so that best-efforts recovery does not rely on CURRENT file at all.
Test plan (dev server):
make check
./db_basic_test --gtest_filter=DBBasicTest.RecoverWithNoCurrentFile
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6970
Reviewed By: anand1976
Differential Revision: D22013990
Pulled By: riversand963
fbshipit-source-id: db552a1868c60ed70e1f7cd252a3a076eb8ea58f
Summary:
Added DB::GetDbSessionId by using the same format and machinery as DB::GetDbIdentity.
The DB Session ID is generated (and therefore, updated) each time a DB object is opened. It is written to the LOG file right after the line of “DB SUMMARY”.
A test for the uniqueness, for different openings and during the same opening, is also added.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6959
Test Plan: Passed make check
Reviewed By: zhichao-cao
Differential Revision: D21951721
Pulled By: gg814
fbshipit-source-id: 958a48a612db49a39998ea703cded45987d3fa8b
Summary:
Persistent cache feature caused rocks db crash on windows. I posted a issue for it, https://github.com/facebook/rocksdb/issues/6919. I found this is because no "persistent_cache_key_prefix" is generated for persistent cache. Looking repo history, "GetUniqueIdFromFile" is not implemented on Windows. So my fix is adding "NewId()" function in "persistent_cache" and using it to generate prefix for persistent cache. In this PR, i also re-enable related test cases defined in "db_test2" and "persistent_cache_test" for windows.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6932
Test Plan:
1. run related test cases in "db_test2" and "persistent_cache_test" on windows and see it passed.
2. manually run db_bench.exe with "read_cache_path" and verified.
Reviewed By: riversand963
Differential Revision: D21911608
Pulled By: cheng-chang
fbshipit-source-id: cdfd938d54a385edbb2836b13aaa1d39b0a6f1c2
Summary:
`Env::LowerThreadPoolCPUPriority` takes a new parameter `CpuPriority` to be able to lower to a specific priority such as `CpuPriority::kIdle`, previously, the priority is always lowered to `CpuPriority::kLow`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6969
Test Plan: unit test `EnvPosixTest::LowerThreadPoolCpuPriority` added to `env_test.cc`.
Reviewed By: siying
Differential Revision: D22011169
Pulled By: cheng-chang
fbshipit-source-id: 568878c24a924912e35cef00c552d4a63431cdf4
Summary:
If `options.wal_recovery_mode == WALRecoveryMode::kPointInTimeRecovery`, RocksDB stops replaying WAL once hitting an error and discards the rest of the WAL. This can lead to data loss if the error occurs at an offset smaller than the last sync'ed offset.
Ideally, RocksDB point-in-time recovery should permit recovery if the error occurs after last synced offset while fail recovery if error occurs before the last synced offset. However, RocksDB does not track the synced offset of WALs. Consequently, RocksDB does not know whether an error occurs before or after the last synced offset. An error can be one of the following.
- WAL record checksum mismatch. This can result from both corruption of synced data and dropping of unsynced data during shutdown. We cannot be sure which one. In order not to defeat the original motivation to permit the latter case, we keep the original behavior of point-in-time WAL recovery.
- IOError. This means the WAL can be bad, an indicator of whole file becoming unavailable, not to mention synced part of the WAL. Therefore, we choose to modify the behavior of point-in-time recovery and fail the database recovery.
Test plan (devserver):
make check
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6963
Reviewed By: ajkr
Differential Revision: D22011083
Pulled By: riversand963
fbshipit-source-id: f9cbf29a37dc5cc40d3fa62f89eed1ad67ca1536
Summary:
Application can ingest SST files with file checksum information, such that during ingestion, DB is able to check data integrity and identify of the SST file. The PR introduces generate_and_verify_file_checksum to IngestExternalFileOption to control if the ingested checksum information should be verified with the generated checksum.
1. If generate_and_verify_file_checksum options is *FALSE*: *1)* if DB does not enable SST file checksum, the checksum information ingested will be ignored; *2)* if DB enables the SST file checksum and the checksum function name matches the checksum function name in DB, we trust the ingested checksum, store it in Manifest. If the checksum function name does not match, we treat that as an error and fail the IngestExternalFile() call.
2. If generate_and_verify_file_checksum options is *TRUE*: *1)* if DB does not enable SST file checksum, the checksum information ingested will be ignored; *2)* if DB enable the SST file checksum, we will use the checksum generator from DB to calculate the checksum for each ingested SST files after they are copied or moved. Then, compare the checksum results with the ingested checksum information: _A)_ if the checksum function name does not match, _verification always report true_ and we store the DB generated checksum information in Manifest. _B)_ if the checksum function name mach, and checksum match, ingestion continues and stores the checksum information in the Manifest. Otherwise, terminate file ingestion and report file corruption.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6891
Test Plan: added unit test, pass make asan_check
Reviewed By: pdillinger
Differential Revision: D21935988
Pulled By: zhichao-cao
fbshipit-source-id: 7b55f486632db467e76d72602218d0658aa7f6ed
Summary:
This saves up to two key comparisons in block seeks. The first key
comparison saved is a redundant key comparison against the restart key
where the linear scan starts. This comparison is saved in all cases
except when the found key is in the first restart interval. The
second key comparison saved is a redundant key comparison against the
restart key where the linear scan ends. This is only saved in cases
where all keys in the restart interval are less than the target
(probability roughly `1/restart_interval`).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6646
Test Plan:
ran a benchmark with mostly default settings and counted key comparisons
before: `user_key_comparison_count = 19399529`
after: `user_key_comparison_count = 18431498`
setup command:
```
$ TEST_TMPDIR=/dev/shm/dbbench ./db_bench -benchmarks=fillrandom,compact -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -max_background_jobs=12 -level_compaction_dynamic_level_bytes=true -num=10000000
```
benchmark command:
```
$ TEST_TMPDIR=/dev/shm/dbbench/ ./db_bench -use_existing_db=true -benchmarks=readrandom -disable_auto_compactions=true -num=10000000 -compression_type=none -reads=1000000 -perf_level=3
```
Reviewed By: pdillinger
Differential Revision: D20849707
Pulled By: ajkr
fbshipit-source-id: 1f01c5cd99ea771fd27974046e37b194f1cdcfac
Summary:
Memory pinned by `pin_l0_filter_and_index_blocks_in_cache` needs to be predictable based on user config. This PR makes sure
we do not pin extra memory for large files generated by intra-L0 (see https://github.com/facebook/rocksdb/issues/6889).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6911
Test Plan: unit test
Reviewed By: siying
Differential Revision: D21835818
Pulled By: ajkr
fbshipit-source-id: a11a088549d06bed8aacc2548d266e5983f0ead4
Summary:
When MultiGet is called with duplicate keys, and the key matches the
largest key in an SST file and the value type is merge, only the first
instance of the duplicate key is returned with correct results. This is
due to the incorrect assumption that if a key in a batch is equal to the
largest key in the file, the next key cannot be present in that file.
Tests:
Add a new unit test
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6953
Reviewed By: cheng-chang
Differential Revision: D21935898
Pulled By: anand1976
fbshipit-source-id: a2cc327a15150e23fd997546ca64d1c33021cb4c
Summary:
Implemented a subcommand of sst_dump called identify, which determines whether a file is an SST file or identifies and lists all the SST files in a directory;
This update also fixes the problem that sst_dump exits with a success state even if target file/directory does not exist/is not an SST file/is empty/is corrupted.
One test is added to sst_dump_test.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6943
Test Plan: Passed make check and a few manual tests
Reviewed By: pdillinger
Differential Revision: D21928985
Pulled By: gg814
fbshipit-source-id: 9a8b48e0cf1a0e96b13f42b690aba8ad981afad3
Summary:
The ```for``` loop in ```VerifyChecksumInBlocks``` only checks ```index_iter->Valid()``` which could be ```false``` either due to reaching the end of the index or, in case of partitioned index, it could be due to a checksum mismatch error when reading a 2nd level index block. Instead of throwing away the index iterator status, we need to return any errors back to the caller.
Tests:
Add a test in block_based_table_reader_test.cc.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6909
Reviewed By: pdillinger
Differential Revision: D21833922
Pulled By: anand1976
fbshipit-source-id: bc778ebf1121dbbdd768689de5183f07a9f0beae
Summary:
When operation on an open file descriptor fails, we should close the file descriptor.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6936
Test Plan: make check
Reviewed By: pdillinger
Differential Revision: D21885458
Pulled By: riversand963
fbshipit-source-id: ba077a76b256a8537f21e22e4ec198f45390bf50
Summary:
DB::OpenForReadOnly will not write anything to the file system (i.e., create directories or files for the DB) unless create_if_missing is true.
This change also fixes some subcommands of ldb, which write to the file system even if the purpose is for readonly.
Two tests for this updated behavior of DB::OpenForReadOnly are also added.
Other minor changes:
1. Updated HISTORY.md to include this API change of DB::OpenForReadOnly;
2. Updated the help information for the put and batchput subcommands of ldb with the option [--create_if_missing];
3. Updated the comment of Env::DeleteDir to emphasize that it returns OK only if the directory to be deleted is empty.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6900
Test Plan: passed make check; also manually tested a few ldb subcommands
Reviewed By: pdillinger
Differential Revision: D21822188
Pulled By: gg814
fbshipit-source-id: 604cc0f0d0326a937ee25a32cdc2b512f9a3be6e
Summary:
The implementation of GetApproximateSizes was inconsistent in
its treatment of the size of non-data blocks of SST files, sometimes
including and sometimes now. This was at its worst with large portion
of table file used by filters and querying a small range that crossed
a table boundary: the size estimate would include large filter size.
It's conceivable that someone might want only to know the size in terms
of data blocks, but I believe that's unlikely enough to ignore for now.
Similarly, there's no evidence the internal function AppoximateOffsetOf
is used for anything other than a one-sided ApproximateSize, so I intend
to refactor to remove redundancy in a follow-up commit.
So to fix this, GetApproximateSizes (and implementation details
ApproximateSize and ApproximateOffsetOf) now consistently include in
their returned sizes a portion of table file metadata (incl filters
and indexes) based on the size portion of the data blocks in range. In
other words, if a key range covers data blocks that are X% by size of all
the table's data blocks, returned approximate size is X% of the total
file size. It would technically be more accurate to attribute metadata
based on number of keys, but that's not computationally efficient with
data available and rarely a meaningful difference.
Also includes miscellaneous comment improvements / clarifications.
Also included is a new approximatesizerandom benchmark for db_bench.
No significant performance difference seen with this change, whether ~700 ops/sec with cache_index_and_filter_blocks and small cache or ~150k ops/sec without cache_index_and_filter_blocks.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6784
Test Plan:
Test added to DBTest.ApproximateSizesFilesWithErrorMargin.
Old code running new test...
[ RUN ] DBTest.ApproximateSizesFilesWithErrorMargin
db/db_test.cc:1562: Failure
Expected: (size) <= (11 * 100), actual: 9478 vs 1100
Other tests updated to reflect consistent accounting of metadata.
Reviewed By: siying
Differential Revision: D21334706
Pulled By: pdillinger
fbshipit-source-id: 6f86870e45213334fedbe9c73b4ebb1d8d611185
Summary:
`IterKey::UpdateInternalKey()` is an error-prone API as it's
incompatible with `IterKey::TrimAppend()`, which is used for
decoding delta-encoded internal keys. This PR stops using it in
`BlockIter`. Instead, it assigns global seqno in a separate `IterKey`'s
buffer when needed. The logic for safely getting a Slice with global
seqno properly assigned is encapsulated in `GlobalSeqnoAppliedKey`.
`BinarySeek()` is also migrated to use this API (previously it ignored
global seqno entirely).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6843
Test Plan:
benchmark setup -- single file DBs, in-memory, no compression. "normal_db"
created by regular flush; "ingestion_db" created by ingesting a file. Both
DBs have same contents.
```
$ TEST_TMPDIR=/dev/shm/normal_db/ ./db_bench -benchmarks=fillrandom,compact -write_buffer_size=10485760000 -disable_auto_compactions=true -compression_type=none -num=1000000
$ ./ldb write_extern_sst ./tmp.sst --db=/dev/shm/ingestion_db/dbbench/ --compression_type=no --hex --create_if_missing < <(./sst_dump --command=scan --output_hex --file=/dev/shm/normal_db/dbbench/000007.sst | awk 'began {print "0x" substr($1, 2, length($1) - 2), "==>", "0x" $5} ; /^Sst file format: block-based/ {began=1}')
$ ./ldb ingest_extern_sst ./tmp.sst --db=/dev/shm/ingestion_db/dbbench/
```
benchmark run command:
```
TEST_TMPDIR=/dev/shm/$DB/ ./db_bench -benchmarks=seekrandom -seek_nexts=10 -use_existing_db=true -cache_index_and_filter_blocks=false -num=1000000 -cache_size=1048576000 -threads=1 -reads=40000000
```
results:
| DB | code | throughput |
|---|---|---|
| normal_db | master | 267.9 |
| normal_db | PR6843 | 254.2 (-5.1%) |
| ingestion_db | master | 259.6 |
| ingestion_db | PR6843 | 250.5 (-3.5%) |
Reviewed By: pdillinger
Differential Revision: D21562604
Pulled By: ajkr
fbshipit-source-id: 937596f836930515da8084d11755e1f247dcb264
Summary:
1. Add a value_size in read options which limits the cumulative value size of keys read in batches. Once the size exceeds read_options.value_size, all the remaining keys are returned with status Abort without further fetching any key.
2. Add a unit test case MultiGetBatchedValueSizeSimple the reads keys from memory and sst files.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6826
Test Plan:
1. make check -j64
2. Add a new unit test case
Reviewed By: anand1976
Differential Revision: D21471483
Pulled By: akankshamahajan15
fbshipit-source-id: dea51b8e76d5d1df38ece8cdb29933b1d798b900
Summary:
If Option.file_checksum_gen_factory is set, rocksdb generates the file checksum during flush and compaction based on the checksum generator created by the factory and store the checksum and function name in vstorage and Manifest.
This PR enable file checksum generation in SstFileWrite and store the checksum and checksum function name in the ExternalSstFileInfo, such that application can use them for other purpose, for example, ingest the file checksum with files in IngestExternalFile().
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6859
Test Plan: add unit test and pass make asan_check.
Reviewed By: ajkr
Differential Revision: D21656247
Pulled By: zhichao-cao
fbshipit-source-id: 78a3570c76031d8832e3d2de3d6c79cdf2b675d0
Summary:
sst_dump can issue many file reads from the file system. This doesn't work well with file systems without a OS cache, especially remote file systems. In order to mitigate this problem, several improvements are done:
1. --readahead_size is added, so that users can specify readahead size when scanning the data.
2. Force a 512KB tail readahead, which prevents three I/Os for footer, meta index and property blocks and hopefully index and filter blocks too.
3. Consoldiate SSTDump's I/Os before opening the file for read. Use the same file prefetch buffer.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6836
Test Plan: Add a test that covers this new feature.
Reviewed By: pdillinger
Differential Revision: D21516607
fbshipit-source-id: 3ae43526286f67b2f4a5bdedfbc92719d579b87e
Summary:
When using ldb, users cannot turn on force consistency check in most commands, while they cannot use checksonsistnecy with --try_load_options. The change fixes both by:
1. checkconsistency now calls OpenDB() so that it gets all the options loading and sanitized options logic
2. use options.check_consistency_checks = true by default, and add a --disable_consistency_checks to turn it off.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6802
Test Plan: Add a new unit test. Some manual tests with corrupted DBs.
Reviewed By: pdillinger
Differential Revision: D21388051
fbshipit-source-id: 8d122732d391b426e3982a1c3232a8e3763ffad0
Summary:
1. Update column_family_memtables_ to point to latest column_family_set in
version_set after recovery.
2. Normalize file paths passed by application so that directories end with '/'
or '\\'.
3. In addition to missing files, corrupted files are also ignored in
best-efforts recovery.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6824
Test Plan: COMPILE_WITH_ASAN=1 make check
Reviewed By: anand1976
Differential Revision: D21463905
Pulled By: riversand963
fbshipit-source-id: c48db8843cc93c8c1c7139c474b64e6f775307d2
Summary:
Delete triggered compaction in universal compaction mode was causing a corruption when scheduled in parallel with other compactions.
1. When num_levels = 1, a file marked for compaction may be picked along with all older files in L0, without checking if any of them are already being compaction. This can cause unpredictable results like resurrection of older versions of keys or deleted keys.
2. When num_levels > 1, a delete triggered compaction would not get scheduled if it overlaps with a running regular compaction. However, the reverse is not true. This is due to the fact that in ```UniversalCompactionBuilder::CalculateSortedRuns```, it assumes that entire sorted runs are picked for compaction and only checks the first file in a sorted run to determine conflicts. This is violated by a delete triggered compaction as it works on a subset of a sorted run.
Fix the bug for num_levels > 1, and disable the feature for now when num_levels = 1. After disabling this feature, files would still get marked for compaction, but no compaction would get scheduled.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6799
Reviewed By: pdillinger
Differential Revision: D21431286
Pulled By: anand1976
fbshipit-source-id: ae9f0bdb1d6ae2f10284847db731c23f43af164a
Summary:
…e range"
Moved it from the wrong section (6.10) to the right section (Unreleased).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6825
Reviewed By: zhichao-cao
Differential Revision: D21464577
Pulled By: ajkr
fbshipit-source-id: a836b4ab10be2464182826f9411c9c424c933b70
Summary:
The error is assigning KeyContext::s to NotFound status in a
table reader for a "not found in this table" case, which skips searching
in later tables, like only a delete should. (The hash search index iterator
is the only one that can return status NotFound even if Valid() == false.)
This was detected by intermittent failure in
MultiThreadedDBTest.MultiThreaded/5, a kHashSearch configuration.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6821
Test Plan: modified existing unit test to reproduce problem
Reviewed By: anand1976
Differential Revision: D21450469
Pulled By: pdillinger
fbshipit-source-id: 7478003684d637dbd491cdac81468041a791be2c
Summary:
We found some files containing nothing but negative range tombstones,
and unsurprisingly their metadata specified a negative range, which made
things crash. Time to add a bit of user input validation.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6788
Reviewed By: zhichao-cao
Differential Revision: D21343719
Pulled By: ajkr
fbshipit-source-id: f1c16e4c3e9fa150958c8c866176632a3206fb74
Summary:
https://github.com/facebook/rocksdb/pull/6807 extends the logic that
identifies and purges obsolete files to blob files handled by RocksDB
itself. In order to prevent that from interfering with the current BlobDB code,
we need to make sure that `BlobDBOptions::blob_dir` is different from
the storage directories used by the base DB. (Note: this is true by default.)
The patch adds a check that explicitly disallows this configuration and
returns `Status::NotSupported` from `BlobDB::Open` in such cases.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6810
Test Plan: Tested using the BlobDB mode of `db_bench`.
Reviewed By: riversand963
Differential Revision: D21412676
Pulled By: ltamasi
fbshipit-source-id: 6630cc7481e48c8bf55d59423b25f14d52ffe681
Summary:
Current impl. of db_stress will abort verification and report failure if
GetLiveFiles() causes a dropped column family to be flushed. This is not
desired.
To fix, this PR makes the following change:
In GetLiveFiles, if flush is triggered and returns
Status::IsColumnFamilyDropped(), then set status to Status::OK().
This is OK because dropped column families will be skipped during the rest of
this function, and valid column families will have their live files returned to
caller.
Test plan (dev server):
make check
./db_stress -ops_per_thread=1000 -get_live_files_one_in=100 -clear_column_family_one_in=100
./db_stress -disable_wal=1 -reopen=0 -ops_per_thread=1000 -get_live_files_one_in=100 -clear_column_family_one_in=100
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6805
Reviewed By: ltamasi
Differential Revision: D21390044
Pulled By: riversand963
fbshipit-source-id: de67846b95a4f1b88aa0a30c3d70c43cc68625b9
Summary:
We are swallowing some file consistency checking failures. This is not expected. We are fixing two cases: DB reopen and manifest dump.
More places are not fixed and need follow-up.
Error from CheckConsistencyForDeletes() is also swallowed, which is not fixed in this PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6793
Test Plan: Add a unit test to cover the reopen case.
Reviewed By: riversand963
Differential Revision: D21366525
fbshipit-source-id: eb438a322237814e8d5125f916a3c6de97f39ded
Andrew Kryczka