Fix data race in accessing recovery_in_prog_
#11950
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@cbi42 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
anand1976
approved these changes
Oct 12, 2023
ajkr
pushed a commit
that referenced
this pull request
Oct 19, 2023
Summary:
We saw the following TSAN stress test failure:
```
WARNING: ThreadSanitizer: data race (pid=17523)
Write of size 1 at 0x7b8c000008b9 by thread T4243 (mutexes: write M0):
#0 rocksdb::ErrorHandler::RecoverFromRetryableBGIOError() fbcode/internal_repo_rocksdb/repo/db/error_handler.cc:742 (db_stress+0x95f954) (BuildId: 35795dfb86ddc9c4f20ddf08a491f24d)
#1 std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (rocksdb::ErrorHandler::*)(), rocksdb::ErrorHandler*>>>::_M_run() fbcode/third-party-buck/platform010/build/libgcc/include/c++/trunk/bits/invoke.h:74 (db_stress+0x95fc2b) (BuildId: 35795dfb86ddc9c4f20ddf08a491f24d)
#2 execute_native_thread_routine /home/engshare/third-party2/libgcc/11.x/src/gcc-11.x/x86_64-facebook-linux/libstdc++-v3/src/c++11/../../../.././libstdc++-v3/src/c++11/thread.cc:82:18 (libstdc++.so.6+0xdf4e4) (BuildId: 452d1cdae868baeeb2fdf1ab140f1c219bf50c6e)
Previous read of size 1 at 0x7b8c000008b9 by thread T22:
#0 rocksdb::DBImpl::SyncClosedLogs(rocksdb::JobContext*, rocksdb::VersionEdit*) fbcode/internal_repo_rocksdb/repo/db/error_handler.h:76 (db_stress+0x84f69c) (BuildId: 35795dfb86ddc9c4f20ddf08a491f24d)
```
This is due to a data race in accessing `recovery_in_prog_`. This PR fixes it by accessing `recovery_in_prog_` under db mutex before calling `SyncClosedLogs()`. I think the original PR #10489 intended to clear the error if it's a recovery flush. So ideally we can also just check flush reason. I plan to keep a safer change in this PR and make that change in the future if needed.
Pull Request resolved: #11950
Test Plan: check future TSAN stress test results.
Reviewed By: anand1976
Differential Revision: D50242255
Pulled By: cbi42
fbshipit-source-id: 0d487948ef9546b038a34460f3bb037f6e5bfc58
cz2h
pushed a commit
to cz2h/rocksdb
that referenced
this pull request
Oct 25, 2023
This reverts commit 6e3429b.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We saw the following TSAN stress test failure:
This is due to a data race in accessing
recovery_in_prog_. This PR fixes it by accessingrecovery_in_prog_under db mutex before callingSyncClosedLogs(). I think the original PR #10489 intended to clear the error if it's a recovery flush. So ideally we can also just check flush reason. I plan to keep a safer change in this PR and make that change in the future if needed.Test plan: check future TSAN stress test results.