Summary:
Added a `CompactionFilter` that is aware of the stress test's expected state. It only drops key versions that are already covered according to the expected state. It is incompatible with snapshots (same as all `CompactionFilter`s), so disables all snapshot-related features when used in the crash test.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6988
Test Plan:
running a minified blackbox crash test
```
$ TEST_TMPDIR=/dev/shm python tools/db_crashtest.py blackbox --max_key=1000000 -write_buffer_size=1048576 -max_bytes_for_level_base=4194304 -target_file_size_base=1048576 -value_size_mult=33 --interval=10 --duration=3600
```
Reviewed By: anand1976
Differential Revision: D22072888
Pulled By: ajkr
fbshipit-source-id: 727b9d7a90d5eab18be0ec6cd5a810712ac13320
Summary:
Currently, `db_stress` tests a randomly picked one of `GetLiveFiles`,
`GetSortedWalFiles`, and `GetCurrentWalFile` with a 1/N chance when the
command line parameter `get_live_files_and_wal_files_one_in` is specified.
The problem is that `GetSortedWalFiles` and `GetCurrentWalFile` are unreliable
in the sense that they can return errors if another thread removes a WAL file
while they are executing (which is a perfectly plausible and legitimate scenario).
The patch splits this command line parameter into three (one for each API),
and changes the crash test script so that only `GetLiveFiles` is tested during
our continuous crash test runs.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6491
Test Plan:
```
make check
python tools/db_crashtest.py whitebox
```
Reviewed By: siying
Differential Revision: D20312200
Pulled By: ltamasi
fbshipit-source-id: e7c3481eddfe3bd3d5349476e34abc9eee5b7dc8
Summary:
When dynamically linking two binaries together, different builds of RocksDB from two sources might cause errors. To provide a tool for user to solve the problem, the RocksDB namespace is changed to a flag which can be overridden in build time.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6433
Test Plan: Build release, all and jtest. Try to build with ROCKSDB_NAMESPACE with another flag.
Differential Revision: D19977691
fbshipit-source-id: aa7f2d0972e1c31d75339ac48478f34f6cfcfb3e
Summary:
Call Transaction::MultiGet from TestMultiGet() in db_stress. We add some Puts/Merges/Deletes into the transaction in order to exercise MultiGetFromBatchAndDB. There is no data verification on read, just check status. There is currently no read data verification in db_stress as it requires synchronization with writes. It needs to be tackled separately.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6227
Test Plan: make crash_test_with_txn
Differential Revision: D19204611
Pulled By: anand1976
fbshipit-source-id: 770d0e30d002e88626c264c58103f1d709bb060c
Summary:
db_stress to execute DB::GetApproximateSizes() with randomized keys and options. Return value is not validated but error will be reported.
Two ways to generate the range keys: (1) two random keys; (2) a small range.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6213
Test Plan: (1) run "make crash_test" for a while; (2) hack the code to ingest some errors to see it is reported.
Differential Revision: D19204665
fbshipit-source-id: 652db36f13bcb5a3bd8fe4a10c0aa22a77a0bce2
Summary:
1. Cover SeekToFirst() and SeekToLast().
2. Try to record the history of iterator operations.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6166
Test Plan: Do some manual changes in the code to cover the failure cases and see the error printing is correct and SeekToFirst() and SeekToLast() sometimes show up.
Differential Revision: D19047079
fbshipit-source-id: 1ed616f919fe4d32c0a021fc37932a7bd3063bcd
Summary:
Add the verification in operateDB to verify GetLiveFiles, GetSortedWalFiles and GetCurrentWalFile. The test will be called every 1 out of N, N is decided by get_live_files_and_wal_files_one_i, whose default is 1000000.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6224
Test Plan: pass db_stress default run.
Differential Revision: D19183358
Pulled By: zhichao-cao
fbshipit-source-id: 20073cf72ede77a3e0d3cf5f28304f1f605d2b1a
Summary:
Currently, db_stress performs verification by calling `VerifyDb()` at the end of test and optionally before tests start. In case of corruption or incorrect result, it will be too late. This PR adds more verification in two ways.
1. For cf consistency test, each test thread takes a snapshot and verifies every N ops. N is configurable via `-verify_db_one_in`. This option is not supported in other stress tests.
2. For cf consistency test, we use another background thread in which a secondary instance periodically tails the primary (interval is configurable). We verify the secondary. Once an error is detected, we terminate the test and report. This does not affect other stress tests.
Test plan (devserver)
```
$./db_stress -test_cf_consistency -verify_db_one_in=0 -ops_per_thread=100000 -continuous_verification_interval=100
$./db_stress -test_cf_consistency -verify_db_one_in=1000 -ops_per_thread=10000 -continuous_verification_interval=0
$make crash_test
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6173
Differential Revision: D19047367
Pulled By: riversand963
fbshipit-source-id: aeed584ad71f9310c111445f34975e5ab47a0615
Summary:
Complete some refactoring called for in https://github.com/facebook/rocksdb/issues/6148. Somehow I got some 'make format' in here for files I didn't change, but that should be OK. (I'm not sure why "hide whitespace changes" doesn't seem to help in review.)
Not addressed in this PR: some operations simply print to stdout rather than failing on discovering a bad status or inconsistency.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6195
Differential Revision: D19131067
Pulled By: pdillinger
fbshipit-source-id: 4f416e6b792023989ef119f385fe122426cb825d
Summary:
CancelAllBackgroundWork() and Close() are frequently used features but we don't cover it in stress test. Simply execute them before closing the DB with 1/2 chance.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6141
Test Plan: Run "db_stress".
Differential Revision: D18900861
fbshipit-source-id: 49b46ccfae120d0f9de3e0543b82fb6d715949d0
Summary:
Right now, in db_stress, compact range is simply executed without any immediate data validation. Add a simply validation which compares hash for all keys within the compact range to stay the same against the same snapshot before and after the compaction.
Also, randomly tune most knobs of CompactRangeOptions.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6140
Test Plan: Run db_stress with "--compact_range_one_in=2000 --compact_range_width=100000000" for a while. Manually ingest some hacky code and observe the error path.
Differential Revision: D18900230
fbshipit-source-id: d96e75bc8c38dd5ec702571ffe7cf5f4ea93ee10
Summary:
db_stress_tool.cc now is a giant file. In order to main it easier to improve and maintain, break it down to multiple source files.
Most classes are turned into their own files. Separate .h and .cc files are created for gflag definiations. Another .h and .cc files are created for some common functions. Some test execution logic that is only loosely related to class StressTest is moved to db_stress_driver.h and db_stress_driver.cc. All the files are located under db_stress_tool/. The directory name is created as such because if we end it with either stress or test, .gitignore will ignore any file under it and makes it prone to issues in developements.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6134
Test Plan: Build under GCC7 with and without LITE on using GNU Make. Build with GCC 4.8. Build with cmake with -DWITH_TOOL=1
Differential Revision: D18876064
fbshipit-source-id: b25d0a7451840f31ac0f5ebb0068785f783fdf7d
