Summary:
The RocksDB iterator is a hierarchy of iterators. MergingIterator maintains a heap of LevelIterators, one for each L0 file and for each non-zero level. The Seek() operation naturally lends itself to parallelization, as it involves positioning every LevelIterator on the correct data block in the correct SST file. It lookups a level for a target key, to find the first key that's >= the target key. This typically involves reading one data block that is likely to contain the target key, and scan forward to find the first valid key. The forward scan may read more data blocks. In order to find the right data block, the iterator may read some metadata blocks (required for opening a file and searching the index).
This flow can be parallelized.
Design: Seek will be called two times under async_io option. First seek will send asynchronous request to prefetch the data blocks at each level and second seek will follow the normal flow and in FilePrefetchBuffer::TryReadFromCacheAsync it will wait for the Poll() to get the results and add the iterator to min_heap.
- Status::TryAgain is passed down from FilePrefetchBuffer::PrefetchAsync to block_iter_.Status indicating asynchronous request has been submitted.
- If for some reason asynchronous request returns error in submitting the request, it will fallback to sequential reading of blocks in one pass.
- If the data already exists in prefetch_buffer, it will return the data without prefetching further and it will be treated as single pass of seek.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9994
Test Plan:
- **Run Regressions.**
```
./db_bench -db=/tmp/prefix_scan_prefetch_main -benchmarks="fillseq" -key_size=32 -value_size=512 -num=5000000 -use_direct_io_for_flush_and_compaction=true -target_file_size_base=16777216
```
i) Previous release 7.0 run for normal prefetching with async_io disabled:
```
./db_bench -use_existing_db=true -db=/tmp/prefix_scan_prefetch_main -benchmarks="seekrandom" -key_size=32 -value_size=512 -num=5000000 -use_direct_reads=true -seek_nexts=327680 -duration=120 -ops_between_duration_checks=1
Initializing RocksDB Options from the specified file
Initializing RocksDB Options from command-line flags
RocksDB: version 7.0
Date: Thu Mar 17 13:11:34 2022
CPU: 24 * Intel Core Processor (Broadwell)
CPUCache: 16384 KB
Keys: 32 bytes each (+ 0 bytes user-defined timestamp)
Values: 512 bytes each (256 bytes after compression)
Entries: 5000000
Prefix: 0 bytes
Keys per prefix: 0
RawSize: 2594.0 MB (estimated)
FileSize: 1373.3 MB (estimated)
Write rate: 0 bytes/second
Read rate: 0 ops/second
Compression: Snappy
Compression sampling rate: 0
Memtablerep: SkipListFactory
Perf Level: 1
------------------------------------------------
DB path: [/tmp/prefix_scan_prefetch_main]
seekrandom : 483618.390 micros/op 2 ops/sec; 338.9 MB/s (249 of 249 found)
```
ii) normal prefetching after changes with async_io disable:
```
./db_bench -use_existing_db=true -db=/tmp/prefix_scan_prefetch_main -benchmarks="seekrandom" -key_size=32 -value_size=512 -num=5000000 -use_direct_reads=true -seek_nexts=327680 -duration=120 -ops_between_duration_checks=1
Set seed to 1652922591315307 because --seed was 0
Initializing RocksDB Options from the specified file
Initializing RocksDB Options from command-line flags
RocksDB: version 7.3
Date: Wed May 18 18:09:51 2022
CPU: 32 * Intel Xeon Processor (Skylake)
CPUCache: 16384 KB
Keys: 32 bytes each (+ 0 bytes user-defined timestamp)
Values: 512 bytes each (256 bytes after compression)
Entries: 5000000
Prefix: 0 bytes
Keys per prefix: 0
RawSize: 2594.0 MB (estimated)
FileSize: 1373.3 MB (estimated)
Write rate: 0 bytes/second
Read rate: 0 ops/second
Compression: Snappy
Compression sampling rate: 0
Memtablerep: SkipListFactory
Perf Level: 1
------------------------------------------------
DB path: [/tmp/prefix_scan_prefetch_main]
seekrandom : 483080.466 micros/op 2 ops/sec 120.287 seconds 249 operations; 340.8 MB/s (249 of 249 found)
```
iii) db_bench with async_io enabled completed succesfully
```
./db_bench -use_existing_db=true -db=/tmp/prefix_scan_prefetch_main -benchmarks="seekrandom" -key_size=32 -value_size=512 -num=5000000 -use_direct_reads=true -seek_nexts=327680 -duration=120 -ops_between_duration_checks=1 -async_io=1 -adaptive_readahead=1
Set seed to 1652924062021732 because --seed was 0
Initializing RocksDB Options from the specified file
Initializing RocksDB Options from command-line flags
RocksDB: version 7.3
Date: Wed May 18 18:34:22 2022
CPU: 32 * Intel Xeon Processor (Skylake)
CPUCache: 16384 KB
Keys: 32 bytes each (+ 0 bytes user-defined timestamp)
Values: 512 bytes each (256 bytes after compression)
Entries: 5000000
Prefix: 0 bytes
Keys per prefix: 0
RawSize: 2594.0 MB (estimated)
FileSize: 1373.3 MB (estimated)
Write rate: 0 bytes/second
Read rate: 0 ops/second
Compression: Snappy
Compression sampling rate: 0
Memtablerep: SkipListFactory
Perf Level: 1
------------------------------------------------
DB path: [/tmp/prefix_scan_prefetch_main]
seekrandom : 553913.576 micros/op 1 ops/sec 120.199 seconds 217 operations; 293.6 MB/s (217 of 217 found)
```
- db_stress with async_io disabled completed succesfully
```
export CRASH_TEST_EXT_ARGS=" --async_io=0"
make crash_test -j
```
I**n Progress**: db_stress with async_io is failing and working on debugging/fixing it.
Reviewed By: anand1976
Differential Revision: D36459323
Pulled By: akankshamahajan15
fbshipit-source-id: abb1cd944abe712bae3986ae5b16704b3338917c
Summary:
This PR implements a coroutine version of batched MultiGet in order to concurrently read from multiple SST files in a level using async IO, thus reducing the latency of the MultiGet. The API from the user perspective is still synchronous and single threaded, with the RocksDB part of the processing happening in the context of the caller's thread. In Version::MultiGet, the decision is made whether to call synchronous or coroutine code.
A good way to review this PR is to review the first 4 commits in order - de773b3, 70c2f70, 10b50e1, and 377a597 - before reviewing the rest.
TODO:
1. Figure out how to build it in CircleCI (requires some dependencies to be installed)
2. Do some stress testing with coroutines enabled
No regression in synchronous MultiGet between this branch and main -
```
./db_bench -use_existing_db=true --db=/data/mysql/rocksdb/prefix_scan -benchmarks="readseq,multireadrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=64 -multiread_batched=true -use_direct_reads=false -duration=60 -ops_between_duration_checks=1 -readonly=true -adaptive_readahead=true -threads=16 -cache_size=10485760000 -async_io=false -multiread_stride=40000 -statistics
```
Branch - ```multireadrandom : 4.025 micros/op 3975111 ops/sec 60.001 seconds 238509056 operations; 2062.3 MB/s (14767808 of 14767808 found)```
Main - ```multireadrandom : 3.987 micros/op 4013216 ops/sec 60.001 seconds 240795392 operations; 2082.1 MB/s (15231040 of 15231040 found)```
More benchmarks in various scenarios are given below. The measurements were taken with ```async_io=false``` (no coroutines) and ```async_io=true``` (use coroutines). For an IO bound workload (with every key requiring an IO), the coroutines version shows a clear benefit, being ~2.6X faster. For CPU bound workloads, the coroutines version has ~6-15% higher CPU utilization, depending on how many keys overlap an SST file.
1. Single thread IO bound workload on remote storage with sparse MultiGet batch keys (~1 key overlap/file) -
No coroutines - ```multireadrandom : 831.774 micros/op 1202 ops/sec 60.001 seconds 72136 operations; 0.6 MB/s (72136 of 72136 found)```
Using coroutines - ```multireadrandom : 318.742 micros/op 3137 ops/sec 60.003 seconds 188248 operations; 1.6 MB/s (188248 of 188248 found)```
2. Single thread CPU bound workload (all data cached) with ~1 key overlap/file -
No coroutines - ```multireadrandom : 4.127 micros/op 242322 ops/sec 60.000 seconds 14539384 operations; 125.7 MB/s (14539384 of 14539384 found)```
Using coroutines - ```multireadrandom : 4.741 micros/op 210935 ops/sec 60.000 seconds 12656176 operations; 109.4 MB/s (12656176 of 12656176 found)```
3. Single thread CPU bound workload with ~2 key overlap/file -
No coroutines - ```multireadrandom : 3.717 micros/op 269000 ops/sec 60.000 seconds 16140024 operations; 139.6 MB/s (16140024 of 16140024 found)```
Using coroutines - ```multireadrandom : 4.146 micros/op 241204 ops/sec 60.000 seconds 14472296 operations; 125.1 MB/s (14472296 of 14472296 found)```
4. CPU bound multi-threaded (16 threads) with ~4 key overlap/file -
No coroutines - ```multireadrandom : 4.534 micros/op 3528792 ops/sec 60.000 seconds 211728728 operations; 1830.7 MB/s (12737024 of 12737024 found) ```
Using coroutines - ```multireadrandom : 4.872 micros/op 3283812 ops/sec 60.000 seconds 197030096 operations; 1703.6 MB/s (12548032 of 12548032 found) ```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9968
Reviewed By: akankshamahajan15
Differential Revision: D36348563
Pulled By: anand1976
fbshipit-source-id: c0ce85a505fd26ebfbb09786cbd7f25202038696
Summary:
1. The latest change of DecideRateLimiterPriority in https://github.com/facebook/rocksdb/pull/9988 is reverted.
2. For https://github.com/facebook/rocksdb/blob/main/db/builder.cc#L345-L349
2.1. Remove `we will regrad this verification as user reads` from the comments.
2.2. `Do not set` the read_options.rate_limiter_priority to Env::IO_USER . Flush should be a background job.
2.3. Update db_rate_limiter_test.cc.
3. In IOOptions, mark `prio` as deprecated for future removal.
4. In `file_system.h`, mark `IOPriority` as deprecated for future removal.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10020
Test Plan: Unit tests.
Reviewed By: ajkr
Differential Revision: D36525317
Pulled By: gitbw95
fbshipit-source-id: 011ba421822f8a124e6d25a2661c4e242df6ad36
Summary:
Start tracking SST unique id in MANIFEST, which is used to verify with
SST properties to make sure the SST file is not overwritten or
misplaced. A DB option `try_verify_sst_unique_id` is introduced to
enable/disable the verification, if enabled, it opens all SST files
during DB-open to read the unique_id from table properties (default is
false), so it's recommended to use it with `max_open_files = -1` to
pre-open the files.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9990
Test Plan: unittests, format-compatible test, mini-crash
Reviewed By: anand1976
Differential Revision: D36381863
Pulled By: jay-zhuang
fbshipit-source-id: 89ea2eb6b35ed3e80ead9c724eb096083eaba63f
Summary:
### Context:
Background compactions and flush generate large reads and writes, and can be long running, especially for universal compaction. In some cases, this can impact foreground reads and writes by users.
### Solution
User, Flush, and Compaction reads share some code path. For this task, we update the rate_limiter_priority in ReadOptions for code paths (e.g. FindTable (mainly in BlockBasedTable::Open()) and various iterators), and eventually update the rate_limiter_priority in IOOptions for FSRandomAccessFile.
**This PR is for the Read path.** The **Read:** dynamic priority for different state are listed as follows:
| State | Normal | Delayed | Stalled |
| ----- | ------ | ------- | ------- |
| Flush (verification read in BuildTable()) | IO_USER | IO_USER | IO_USER |
| Compaction | IO_LOW | IO_USER | IO_USER |
| User | User provided | User provided | User provided |
We will respect the read_options that the user provided and will not set it.
The only sst read for Flush is the verification read in BuildTable(). It claims to be "regard as user read".
**Details**
1. Set read_options.rate_limiter_priority dynamically:
- User: Do not update the read_options. Use the read_options that the user provided.
- Compaction: Update read_options in CompactionJob::ProcessKeyValueCompaction().
- Flush: Update read_options in BuildTable().
2. Pass the rate limiter priority to FSRandomAccessFile functions:
- After calling the FindTable(), read_options is passed through GetTableReader(table_cache.cc), BlockBasedTableFactory::NewTableReader(block_based_table_factory.cc), and BlockBasedTable::Open(). The Open() needs some updates for the ReadOptions variable and the updates are also needed for the called functions, including PrefetchTail(), PrepareIOOptions(), ReadFooterFromFile(), ReadMetaIndexblock(), ReadPropertiesBlock(), PrefetchIndexAndFilterBlocks(), and ReadRangeDelBlock().
- In RandomAccessFileReader, the functions to be updated include Read(), MultiRead(), ReadAsync(), and Prefetch().
- Update the downstream functions of NewIndexIterator(), NewDataBlockIterator(), and BlockBasedTableIterator().
### Test Plans
Add unit tests.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9996
Reviewed By: anand1976
Differential Revision: D36452483
Pulled By: gitbw95
fbshipit-source-id: 60978204a4f849bb9261cb78d9bc1cb56d6008cf
Summary:
### Context:
Background compactions and flush generate large reads and writes, and can be long running, especially for universal compaction. In some cases, this can impact foreground reads and writes by users.
From the RocksDB perspective, there can be two kinds of rate limiters, the internal (native) one and the external one.
- The internal (native) rate limiter is introduced in [the wiki](https://github.com/facebook/rocksdb/wiki/Rate-Limiter). Currently, only IO_LOW and IO_HIGH are used and they are set statically.
- For the external rate limiter, in FSWritableFile functions, IOOptions is open for end users to set and get rate_limiter_priority for their own rate limiter. Currently, RocksDB doesn’t pass the rate_limiter_priority through IOOptions to the file system.
### Solution
During the User Read, Flush write, Compaction read/write, the WriteController is used to determine whether DB writes are stalled or slowed down. The rate limiter priority (Env::IOPriority) can be determined accordingly. We decided to always pass the priority in IOOptions. What the file system does with it should be a contract between the user and the file system. We would like to set the rate limiter priority at file level, since the Flush/Compaction job level may be too coarse with multiple files and block IO level is too granular.
**This PR is for the Write path.** The **Write:** dynamic priority for different state are listed as follows:
| State | Normal | Delayed | Stalled |
| ----- | ------ | ------- | ------- |
| Flush | IO_HIGH | IO_USER | IO_USER |
| Compaction | IO_LOW | IO_USER | IO_USER |
Flush and Compaction writes share the same call path through BlockBaseTableWriter, WritableFileWriter, and FSWritableFile. When a new FSWritableFile object is created, its io_priority_ can be set dynamically based on the state of the WriteController. In WritableFileWriter, before the call sites of FSWritableFile functions, WritableFileWriter::DecideRateLimiterPriority() determines the rate_limiter_priority. The options (IOOptions) argument of FSWritableFile functions will be updated with the rate_limiter_priority.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9988
Test Plan: Add unit tests.
Reviewed By: anand1976
Differential Revision: D36395159
Pulled By: gitbw95
fbshipit-source-id: a7c82fc29759139a1a07ec46c37dbf7e753474cf
Summary:
128 bits should suffice almost always and for tracking in manifest.
Note that this changes the output of sst_dump --show_properties to only show 128 bits.
Also introduces InternalUniqueIdToHumanString for presenting internal IDs for debugging purposes.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10009
Test Plan: unit tests updated
Reviewed By: jay-zhuang
Differential Revision: D36458189
Pulled By: pdillinger
fbshipit-source-id: 93ebc4a3b6f9c73ee154383a1f8b291a5d6bbef5
Summary:
**Context:**
Previous PR https://github.com/facebook/rocksdb/pull/9748, https://github.com/facebook/rocksdb/pull/9073, https://github.com/facebook/rocksdb/pull/8428 added separate flag for each charged memory area. Such API design is not scalable as we charge more and more memory areas. Also, we foresee an opportunity to consolidate this feature with other cache usage related features such as `cache_index_and_filter_blocks` using `CacheEntryRole`.
Therefore we decided to consolidate all these flags with `CacheUsageOptions cache_usage_options` and this PR serves as the first step by consolidating memory-charging related flags.
**Summary:**
- Replaced old API reference with new ones, including making `kCompressionDictionaryBuildingBuffer` opt-out and added a unit test for that
- Added missing db bench/stress test for some memory charging features
- Renamed related test suite to indicate they are under the same theme of memory charging
- Refactored a commonly used mocked cache component in memory charging related tests to reduce code duplication
- Replaced the phrases "memory tracking" / "cache reservation" (other than CacheReservationManager-related ones) with "memory charging" for standard description of this feature.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9926
Test Plan:
- New unit test for opt-out `kCompressionDictionaryBuildingBuffer` `TEST_F(ChargeCompressionDictionaryBuildingBufferTest, Basic)`
- New unit test for option validation/sanitization `TEST_F(CacheUsageOptionsOverridesTest, SanitizeAndValidateOptions)`
- CI
- db bench (in case querying new options introduces regression) **+0.5% micros/op**: `TEST_TMPDIR=/dev/shm/testdb ./db_bench -benchmarks=fillseq -db=$TEST_TMPDIR -charge_compression_dictionary_building_buffer=1(remove this for comparison) -compression_max_dict_bytes=10000 -disable_auto_compactions=1 -write_buffer_size=100000 -num=4000000 | egrep 'fillseq'`
#-run | (pre-PR) avg micros/op | std micros/op | (post-PR) micros/op | std micros/op | change (%)
-- | -- | -- | -- | -- | --
10 | 3.9711 | 0.264408 | 3.9914 | 0.254563 | 0.5111933721
20 | 3.83905 | 0.0664488 | 3.8251 | 0.0695456 | **-0.3633711465**
40 | 3.86625 | 0.136669 | 3.8867 | 0.143765 | **0.5289363078**
- db_stress: `python3 tools/db_crashtest.py blackbox -charge_compression_dictionary_building_buffer=1 -charge_filter_construction=1 -charge_table_reader=1 -cache_size=1` killed as normal
Reviewed By: ajkr
Differential Revision: D36054712
Pulled By: hx235
fbshipit-source-id: d406e90f5e0c5ea4dbcb585a484ad9302d4302af
Summary:
PR https://github.com/facebook/rocksdb/issues/9888 started to enforce the contract of single delete described in https://github.com/facebook/rocksdb/wiki/Single-Delete.
For some of existing use cases, it is desirable to have a transition during which compaction will not fail
if the contract is violated. Therefore, we add a temporary option `enforce_single_del_contracts` to allow
application to opt out from this new strict behavior. Once transition completes, the flag can be set to `true` again.
In a future release, the option will be removed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9983
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D36333672
Pulled By: riversand963
fbshipit-source-id: dcb703ea0ed08076a1422f1bfb9914afe3c2caa2
Summary:
In one path of BlockBasedTable::MultiGet(), Next() is directly called after calling Seek() against the index iterator. This might cause crash if an I/O error happens in Seek().
The bug is discovered in crash test.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9993
Test Plan: See existing CI tests pass.
Reviewed By: anand1976
Differential Revision: D36381758
fbshipit-source-id: a11e0aa48dcee168c2554c33b532646ffdb68877
Summary:
Add methods to set the various functions (Parse, Serialize, Equals) to the OptionTypeInfo. These methods simplify the number of constructors required for OptionTypeInfo and make the code a little clearer.
Add functions to the OptionTypeInfo for Prepare and Validate. These methods allow types other than Configurable and Customizable to have Prepare and Validate logic. These methods could be used by an option to guarantee that its settings were in a range or that a value was initialized.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9411
Reviewed By: pdillinger
Differential Revision: D36174849
Pulled By: mrambacher
fbshipit-source-id: 72517d8c6bab4723788a4c1a9e16590bff870125
Summary:
PR 9929 adds a new CompactionFilter::Decision, i.e.
kRemoveWithSingleDelete so that CompactionFilter can indicate to
CompactionIterator that a PUT can only be removed with SD. However, how
CompactionIterator handles such a key is implementation detail which
should not be implied in the public API. In fact,
such a PUT can just be dropped. This is an optimization which we will apply in the near future.
Discussion thread: https://github.com/facebook/rocksdb/pull/9929#discussion_r863198964
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9951
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D36156590
Pulled By: riversand963
fbshipit-source-id: 7b7d01f47bba4cad7d9cca6ca52984f27f88b372
Summary:
If the DB path is specified, the user would expect ldb loads the
options from the path, but it's not:
```
$ ldb list_live_files_metadata --db=`pwd`
```
Default `try_load_options` to true in that case. The user can still
disable that by:
```
$ ldb list_live_files_metadata --db=`pwd` --try_load_options=false
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9937
Test Plan:
`ldb list_live_files_metadata --db=`pwd`` is able to work for
a db generated with different options.num_levels.
Reviewed By: ajkr
Differential Revision: D36106708
Pulled By: jay-zhuang
fbshipit-source-id: 2732fdc027a4d172436b2c9b6a9787b56b10c710
Summary:
When compaction filter determines that a key should be removed, it updates the internal key's type
to `Delete`. If this internal key is preserved in current compaction but seen by a later compaction
together with `SingleDelete`, it will cause compaction iterator to return Corruption.
To fix the issue, compaction filter should return more information in addition to the intention of removing
a key. Therefore, we add a new `kRemoveWithSingleDelete` to `CompactionFilter::Decision`. Seeing
`kRemoveWithSingleDelete`, compaction iterator will update the op type of the internal key to `kTypeSingleDelete`.
In addition, I updated db_stress_shared_state.[cc|h] so that `no_overwrite_ids_` becomes `const`. It is easier to
reason about thread-safety if accessed from multiple threads. This information is passed to `PrepareTxnDBOptions()`
when calling from `Open()` so that we can set up the rollback deletion type callback for transactions.
Finally, disable compaction filter for multiops_txn because the key removal logic of `DbStressCompactionFilter` does
not quite work with `MultiOpsTxnsStressTest`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9929
Test Plan:
make check
make crash_test
make crash_test_with_txn
Reviewed By: anand1976
Differential Revision: D36069678
Pulled By: riversand963
fbshipit-source-id: cedd2f1ba958af59ad3916f1ba6f424307955f92
Summary:
Enforce the contract of SingleDelete so that they are not mixed with
Delete for the same key. Otherwise, it will lead to undefined behavior.
See https://github.com/facebook/rocksdb/wiki/Single-Delete#notes.
Also fix unit tests and write-unprepared.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9888
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D35837817
Pulled By: riversand963
fbshipit-source-id: acd06e4dcba8cb18df92b44ed18c57e10e5a7635
Summary:
When MultiGet() determines that multiple query keys can be
served by examining the same data block in block cache (one Lookup()),
each PinnableSlice referring to data in that data block needs to hold
on to the block in cache so that they can be released at arbitrary
times by the API user. Historically this is accomplished with extra
calls to Ref() on the Handle from Lookup(), with each PinnableSlice
cleanup calling Release() on the Handle, but this creates extra
contention on the block cache for the extra Ref()s and Release()es,
especially because they hit the same cache shard repeatedly.
In the case of merge operands (possibly more cases?), the problem was
compounded by doing an extra Ref()+eventual Release() for each merge
operand for a key reusing a block (which could be the same key!), rather
than one Ref() per key. (Note: the non-shared case with `biter` was
already one per key.)
This change optimizes MultiGet not to rely on these extra, contentious
Ref()+Release() calls by instead, in the shared block case, wrapping
the cache Release() cleanup in a refcounted object referenced by the
PinnableSlices, such that after the last wrapped reference is released,
the cache entry is Release()ed. Relaxed atomic refcounts should be
much faster than mutex-guarded Ref() and Release(), and much less prone
to a performance cliff when MultiGet() does a lot of block sharing.
Note that I did not use std::shared_ptr, because that would require an
extra indirection object (shared_ptr itself new/delete) in order to
associate a ref increment/decrement with a Cleanable cleanup entry. (If
I assumed it was the size of two pointers, I could do some hackery to
make it work without the extra indirection, but that's too fragile.)
Some details:
* Fixed (removed) extra block cache tracing entries in cases of cache
entry reuse in MultiGet, but it's likely that in some other cases traces
are missing (XXX comment inserted)
* Moved existing implementations for cleanable.h from iterator.cc to
new cleanable.cc
* Improved API comments on Cleanable
* Added a public SharedCleanablePtr class to cleanable.h in case others
could benefit from the same pattern (potentially many Cleanables and/or
smart pointers referencing a shared Cleanable)
* Add a typedef for MultiGetContext::Mask
* Some variable renaming for clarity
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9899
Test Plan:
Added unit tests for SharedCleanablePtr.
Greatly enhanced ability of existing tests to detect cache use-after-free.
* Release PinnableSlices from MultiGet as they are read rather than in
bulk (in db_test_util wrapper).
* In ASAN build, default to using a trivially small LRUCache for block_cache
so that entries are immediately erased when unreferenced. (Updated two
tests that depend on caching.) New ASAN testsuite running time seems
OK to me.
If I introduce a bug into my implementation where we skip the shared
cleanups on block reuse, ASAN detects the bug in
`db_basic_test *MultiGet*`. If I remove either of the above testing
enhancements, the bug is not detected.
Consider for follow-up work: manipulate or randomize ordering of
PinnableSlice use and release from MultiGet db_test_util wrapper. But in
typical cases, natural ordering gives pretty good functional coverage.
Performance test:
In the extreme (but possible) case of MultiGetting the same or adjacent keys
in a batch, throughput can improve by an order of magnitude.
`./db_bench -benchmarks=multireadrandom -db=/dev/shm/testdb -readonly -num=5 -duration=10 -threads=20 -multiread_batched -batch_size=200`
Before ops/sec, num=5: 1,384,394
Before ops/sec, num=500: 6,423,720
After ops/sec, num=500: 10,658,794
After ops/sec, num=5: 16,027,257
Also note that previously, with high parallelism, having query keys
concentrated in a single block was worse than spreading them out a bit. Now
concentrated in a single block is faster than spread out, which is hopefully
consistent with natural expectation.
Random query performance: with num=1000000, over 999 x 10s runs running before & after simultaneously (each -threads=12):
Before: multireadrandom [AVG 999 runs] : 1088699 (± 7344) ops/sec; 120.4 (± 0.8 ) MB/sec
After: multireadrandom [AVG 999 runs] : 1090402 (± 7230) ops/sec; 120.6 (± 0.8 ) MB/sec
Possibly better, possibly in the noise.
Reviewed By: anand1976
Differential Revision: D35907003
Pulled By: pdillinger
fbshipit-source-id: bbd244d703649a8ca12d476f2d03853ed9d1a17e
Summary:
Left HISTORY.md and unit tests.
Added a new unit test to repro the corruption scenario that this PR fixes, and HISTORY.md line for that.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9906
Reviewed By: riversand963
Differential Revision: D35940093
Pulled By: ajkr
fbshipit-source-id: 9816f99e1ce405ba36f316beb4f6378c37c8c86b
Summary:
Add stats PREFETCHED_BYTES_DISCARDED and POLL_WAIT_MICROS.
PREFETCHED_BYTES_DISCARDED records number of prefetched bytes discarded by
FilePrefetchBuffer. POLL_WAIT_MICROS records the time taken by underling
file_system Poll API.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9845
Test Plan: Update existing tests
Reviewed By: anand1976
Differential Revision: D35909694
Pulled By: akankshamahajan15
fbshipit-source-id: e009ef940bb9ed72c9446f5529095caabb8a1e36
Summary:
This PR does not affect write-committed.
Add a member, `rollback_deletion_type_callback` to TransactionDBOptions
so that a write-prepared transaction, when rolling back, can call this
callback to decide if a `Delete` or `SingleDelete` should be used to
cancel a prior `Put` written to the database during prepare phase.
The purpose of this PR is to prevent mixing `Delete` and `SingleDelete`
for the same key, causing undefined behaviors. Without this PR, the
following can happen:
```
// The application always issues SingleDelete when deleting keys.
txn1->Put('a');
txn1->Prepare(); // writes to memtable and potentially gets flushed/compacted to Lmax
txn1->Rollback(); // inserts DELETE('a')
txn2->Put('a');
txn2->Commit(); // writes to memtable and potentially gets flushed/compacted
```
In the database, we may have
```
L0: [PUT('a', s=100)]
L1: [DELETE('a', s=90)]
Lmax: [PUT('a', s=0)]
```
If a compaction compacts L0 and L1, then we have
```
L1: [PUT('a', s=100)]
Lmax: [PUT('a', s=0)]
```
If a future transaction issues a SingleDelete, we have
```
L0: [SD('a', s=110)]
L1: [PUT('a', s=100)]
Lmax: [PUT('a', s=0)]
```
Then, a compaction including L0, L1 and Lmax leads to
```
Lmax: [PUT('a', s=0)]
```
which is incorrect.
Similar bugs reported and addressed in
https://github.com/cockroachdb/pebble/issues/1255. Based on our team's
current priority, we have decided to take this approach for now. We may
come back and revisit in the future.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9873
Test Plan: make check
Reviewed By: ltamasi
Differential Revision: D35762170
Pulled By: riversand963
fbshipit-source-id: b28d56eefc786b53c9844b9ef4a7807acdd82c8d
Summary:
... by filling out remaining testing hole: handling of
db_pathsi+cf_paths. (Note that while GetLiveFilesStorageInfo works
with db_paths / cf_paths, Checkpoint and BackupEngine do not and
are marked appropriately.)
Also improved comments for "live files" APIs, and grouped them
together in db.h.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9868
Test Plan: Adding to existing unit tests
Reviewed By: jay-zhuang
Differential Revision: D35752254
Pulled By: pdillinger
fbshipit-source-id: c70eb67748fad61826e2f554b674638700abefb2
Summary:
We don't really have a mechanism for internal-only release
notes, so adding this to the standard release notes. For picking into
7.2 release.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9872
Test Plan: release note only
Reviewed By: jay-zhuang
Differential Revision: D35761307
Pulled By: pdillinger
fbshipit-source-id: 5d1932767fff48456323df948604dbb956ac27b2
Summary:
Add a merge operator that allows users to register specific aggregation function so that they can does aggregation based per key using different aggregation types.
See comments of function CreateAggMergeOperator() for actual usage.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9780
Test Plan: Add a unit test to coverage various cases.
Reviewed By: ltamasi
Differential Revision: D35267444
fbshipit-source-id: 5b02f31c4f3e17e96dd4025cdc49fca8c2868628
Summary:
This new options allows application to specify that files must be
ingested to bottommost level, otherwise the ingestion will fail instead
of silently ingesting to a non-bottommost level.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9849
Test Plan: make check
Reviewed By: ajkr
Differential Revision: D35680307
Pulled By: riversand963
fbshipit-source-id: 01cf54ef6c76198f7654dc06b5544631dea1be1e
Summary:
Make `DB::GetUpdatesSince` return early if told to scan WALs generated by transactions
with write-prepared or write-unprepared policies (`seq_per_batch` is true), as indicated by
API comment.
Also add checks to `TransactionLogIterator` to clarify some conditions.
No API change.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9459
Test Plan:
make check
Closing https://github.com/facebook/rocksdb/issues/1565
Reviewed By: akankshamahajan15
Differential Revision: D33821243
Pulled By: riversand963
fbshipit-source-id: c8b155d020ce0980e2d3b3b1da40b96e65b48d79
Summary:
This gives users the ability to examine the map populated by `GetMapProperty()` with property `kBlockCacheEntryStats`. It also sets us up for a possible future where cache reservations are configured according to `CacheEntryRole`s rather than flags coupled to roles.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9838
Test Plan:
- migrated test DBBlockCacheTest.CacheEntryRoleStats to use this API. That test verifies some of the contents are as expected
- added a DBPropertiesTest to verify the public map keys are present, and nothing else
Reviewed By: hx235
Differential Revision: D35629493
Pulled By: ajkr
fbshipit-source-id: 5c4356b8560e85d1f881fd32c44c15960b02fc68
Summary:
This information has been already available as part of the `rocksdb.blob-stats`
string property. The patch adds a dedicated integer property to make it easier
to surface this information in monitoring systems.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9835
Test Plan: `make check`
Reviewed By: riversand963
Differential Revision: D35619495
Pulled By: ltamasi
fbshipit-source-id: 03fb0b228aa27d3859a1e3783bcb7eca095607f8
Summary:
So the user is able to set event listener on the compactor
side.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9821
Test Plan: unittest added
Reviewed By: ajkr
Differential Revision: D35485388
Pulled By: jay-zhuang
fbshipit-source-id: 669d8a3aaee012b75b940470306756c03ffa09b2
Summary:
1) In case of non-TransactionDB and avoid_flush_during_recovery = true, RocksDB won't
flush the data from WAL to L0 for all column families if possible. As a
result, not all column families can increase their log_numbers, and
min_log_number_to_keep won't change.
2) For transaction DB (.allow_2pc), even with the flush, there may be old WAL files that it must not delete because they can contain data of uncommitted transactions and min_log_number_to_keep won't change.
If we persist a new MANIFEST with
advanced log_numbers for some column families, then during a second
crash after persisting the MANIFEST, RocksDB will see some column
families' log_numbers larger than the corrupted wal, and the "column family inconsistency" error will be hit, causing recovery to fail.
As a solution,
1. the corrupted WALs whose numbers are larger than the
corrupted wal and smaller than the new WAL will be moved to archive folder.
2. Currently, RocksDB DB::Open() may creates and writes to two new MANIFEST files even before recovery succeeds. This PR buffers the edits in a structure and writes to a new MANIFEST after recovery is successful
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9634
Test Plan:
1. Added new unit tests
2. make crast_test -j
Reviewed By: riversand963
Differential Revision: D34463666
Pulled By: akankshamahajan15
fbshipit-source-id: e233d3af0ed4e2028ca0cf051e5a334a0fdc9d19
Summary:
Currently async prefetching is enabled for implicit internal auto readahead in FilePrefetchBuffer if `ReadOptions.async_io` is set. This PR enables async prefetching for `ReadOptions.readahead_size` when `ReadOptions.async_io` is set true.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9827
Test Plan: Update unit test
Reviewed By: anand1976
Differential Revision: D35552129
Pulled By: akankshamahajan15
fbshipit-source-id: d9f9a96672852a591375a21eef15355cf3289f5c
Summary:
Update stats in random_access_file_reader for Read and
ReadAsync API to take into account the read latency for async
prefetching.
It also fixes ERROR_HANDLER_AUTORESUME_RETRY_COUNT stat whose value was
incorrect in portal.h
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9810
Test Plan: Update unit test
Reviewed By: anand1976
Differential Revision: D35433081
Pulled By: akankshamahajan15
fbshipit-source-id: aeec3901270e58a003ce6b5214bd25ddcb3a12a9
Summary:
For write-prepared/write-unprepared transactions,
GetCommitTimeWriteBatch() can be used only if the transaction is started
with `TransactionOptions::use_only_the_last_commit_time_batch_for_recovery` set
to true. Otherwise, it is possible that multiple uncommitted versions of the
same key exist in the database. During bottommost compaction, RocksDB may
set the sequence numbers of both to zero once they become committed, causing
output SST file to have two identical internal keys.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9794
Test Plan:
make check
pay special attention to the following
```
transaction_test --gtest_filter=MySQLStyleTransactionTest/MySQLStyleTransactionTest.TransactionStressTest/*
```
Reviewed By: lth
Differential Revision: D35327214
Pulled By: riversand963
fbshipit-source-id: 3bae00a28359c10e96e4c6f676d20de5610d8a0f
Summary:
If FilePrefetchBuffer object is destroyed and then later Poll() calls callback on object which has been destroyed, it gives segfault on accessing destroyed object. It was caught after adding unit tests that tests Posix implementation of ReadAsync and Poll APIs.
This PR also updates and fixes existing IOURing tests which were not running locally because RocksDbIOUringEnable function wasn't defined and IOUring was disabled for those tests
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9777
Test Plan: Added new unit test
Reviewed By: anand1976
Differential Revision: D35254002
Pulled By: akankshamahajan15
fbshipit-source-id: 68e80054ffb14ae25c255920ebc6548ca5f130a1
Summary:
min_log_number_to_keep denotes that the WALs whose numbers are below
this value **will** be deleted by RocksDB.
delete_wals_before will be used by RocksDB if
track_and_verify_wals_in_manifest is set to true. During recovery,
RocksDB uses the info encoded in delete_wals_before to reconstruct its
knowledge about what WALs to expect existing.
If these two tags are not encoded in the same VersionEdit, then it's
possible for min_log_number_to_keep=100 to exist, but
delete_wals_before=100 to be lost due to power failure. Subsequent
recovery will delete 99.log. If the db crashes again, the following
recovery will expect to see 99.log since there is no
delete_wals_before=100 in the MANIFEST, but the WAL is already deleted.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9766
Test Plan:
First of all, make check.
Second, format compatibility.
SHORT_TEST=1 ./tools/check_format_compatible.sh
Reviewed By: ltamasi
Differential Revision: D35203623
Pulled By: riversand963
fbshipit-source-id: 45623fc4b4b50d299d5e0f9559a3a4c5e9522c8f
Summary:
In making `SstFileMetaData` inherit from `FileStorageInfo`, I
overlooked setting some `FileStorageInfo` fields when then default
`SstFileMetaData()` ctor is used. This affected `GetLiveFilesMetaData()`.
Also removed some buggy `static_cast<size_t>`
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9769
Test Plan: Updated tests
Reviewed By: jay-zhuang
Differential Revision: D35220383
Pulled By: pdillinger
fbshipit-source-id: 05b4ee468258dbd3699517e1124838bf405fe7f8
Summary:
After commit [d642c60](d642c60bdc), the stats `READ_BLOCK_COMPACTION_MICROS` cannot record any compaction read duration, and it always report zero.
This PR targets to distinguish `READ_BLOCK_COMPACTION_MICROS` with `READ_BLOCK_GET_MICROS` so that `READ_BLOCK_COMPACTION_MICROS` could record the correct stats.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9722
Reviewed By: ajkr
Differential Revision: D35021870
Pulled By: jay-zhuang
fbshipit-source-id: f1a804994265e51465de64c2a08f2e0eeb6fc5a3
Summary:
Although ColumnFamilySet comments say that DB mutex can be
freed during iteration, as long as you hold a ref while releasing DB
mutex, this is not quite true because UnrefAndTryDelete might delete cfd
right before it is needed to get ->next_ for the next iteration of the
loop.
This change solves the problem by making a wrapper class that makes such
iteration easier while handling the tricky details of UnrefAndTryDelete
on the previous cfd only after getting next_ in operator++.
FreeDeadColumnFamilies should already have been obsolete; this removes
it for good. Similarly, ColumnFamilySet::iterator doesn't need to check
for cfd with 0 refs, because those are immediately deleted.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9730
Test Plan:
was reported with ASAN on unit tests like
DBLogicalBlockSizeCacheTest.CreateColumnFamily (very rare); keep watching
Reviewed By: ltamasi
Differential Revision: D35038143
Pulled By: pdillinger
fbshipit-source-id: 0a5478d5be96c135343a00603711b7df43ae19c9