Fix stress test failure "Corruption: checksum mismatch" or "Iterator Diverged" with async_io enabled #10032
Conversation
akankshamahajan15
force-pushed
the
async_io_bug_fix
branch
from
26a350f
to
318ca5a
Compare
|
@akankshamahajan15 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
| // are not sequential and Non sequential reads will be handled there. | ||
| if (!enable_ || (offset < bufs_[curr_].offset_ && async_io_ == false)) { | ||
| // requested bytes. | ||
| if (bufs_[curr_].buffer_.CurrentSize() > 0 && offset < bufs_[curr_].offset_ && |
There was a problem hiding this comment.
This check basically fixes failure.
akankshamahajan15
changed the title
akankshamahajan15
force-pushed
the
async_io_bug_fix
branch
from
318ca5a
to
8338cb9
Compare
|
@akankshamahajan15 has updated the pull request. You must reimport the pull request before landing. |
|
@akankshamahajan15 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags:
akankshamahajan15
force-pushed
the
async_io_bug_fix
branch
from
8338cb9
to
12c936d
Compare
|
@akankshamahajan15 has updated the pull request. You must reimport the pull request before landing. |
akankshamahajan15
force-pushed
the
async_io_bug_fix
branch
from
12c936d
to
bdb5201
Compare
|
@akankshamahajan15 has updated the pull request. You must reimport the pull request before landing. |
akankshamahajan15
changed the title
|
@akankshamahajan15 has updated the pull request. You must reimport the pull request before landing. |
1 similar comment
|
@akankshamahajan15 has updated the pull request. You must reimport the pull request before landing. |
|
@akankshamahajan15 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
Summary: In case of non sequential reads with
async_io,FilePRefetchBuffer::TryReadFromCacheAsynccan be called for previous blocks withoffset < bufs_[curr_].offset_which wasn't handled correctly resulting wrong data being returned from buffer.Since
FilePRefetchBuffer::PrefetchAsynccan be called for any data block, it setsprev_len_to 0 indicatingFilePRefetchBuffer::TryReadFromCacheAsyncto go for the prefetching even though offset < bufs_[curr_].offset_ This is because async prefetching is always done in second buffer (to avoid mutex) even though curr_ is empty leading to offset < bufs_[curr_].offset_ in some cases.If prev_len_ is non zero then
TryReadFromCacheAsyncreturns false ifoffset < bufs_[curr_].offset_ && prev_len != 0indicating reads are not sequential and previous call wasn't PrefetchAsync.FilePRefetchBuffer::TryReadFromCacheAsyncas it was getting complicated covering different scenarios based onasync_ioenabled/disabled. Iffor_compactionis set true, it now callsFilePRefetchBufferTryReadFromCachefollowing synchronous flow as before. Its decided in BlockFetcher.ccTest Plan:
make crash_test -j completed successfully locally