release-26.1.1-rc: vecann: fix concurrent download race in dataset cache#164769
Merged
mw5h merged 1 commit intocockroachdb:release-26.1.1-rcfrom Mar 3, 2026
Merged
Conversation
The vecindex roachtest has been failing 100% of the time on master since
the prefix=0 and prefix=3 test variants run concurrently and both
download the same dataset to the same cache directory. The previous code
used a fixed temp file path (destPath + ".zip") for the download, so
concurrent downloaders would clobber each other's writes, producing
either a corrupt zip ("unexpected EOF" during extraction) or a missing
file ("no such file or directory" when one process's defer cleanup
deletes the file before the other can read it).
Fix this by using os.CreateTemp for both the downloaded zip and the
extracted output, giving each concurrent downloader its own unique temp
files. The extracted file is then installed at the destination via atomic
os.Rename, ensuring the cached file is either absent or complete — never
a truncated partial write.
This also removes the ResetCache field from DatasetLoader entirely. It
was only set to true in the vecindex roachtest (added in 2794f07 as
a workaround for corrupted cache files persisting across runs). With
atomic extraction, the cache can never contain a truncated file, so the
workaround is no longer needed. Since no callers remain, the field and
all associated checks are removed.
Fixes: cockroachdb#163471
Fixes: cockroachdb#159333
Release note: None
Co-Authored-By: roachdev-claude <roachdev-claude-bot@cockroachlabs.com>
|
Thanks for opening a backport. Before merging, please confirm that the change does not break backwards compatibility and otherwise complies with the backport policy. Include a brief release justification in the PR description explaining why the backport is appropriate. All backports must be reviewed by the TL for the owning area. While the stricter LTS policy does not yet apply, please exercise judgment and consider gating non-critical changes behind a disabled-by-default feature flag when appropriate. |
blathers-crl
bot
added
backport
mw5h
requested review from
herkolategan and
srosenberg
and removed request for
a team
Member
|
This change is |
yuzefovich
approved these changes
Mar 3, 2026
Member
yuzefovich
left a comment
yuzefovich
left a comment
There was a problem hiding this comment.
@yuzefovich reviewed 2 files and all commit messages.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on herkolategan and srosenberg).
mw5h
added
the
backport-test-only
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Backport 1/1 commits from #164563.
/cc @cockroachdb/release
Summary
vecindex/dbpedia-100k/nodes=3/prefix=0on master.The
prefix=0andprefix=3test variants run concurrently and both downloadthe same dataset to the same cache directory using a fixed temp file path,
causing concurrent writers to corrupt each other's data.
downloadAndUnzipto useos.CreateTempfor unique temp files perdownloader, with atomic
os.Renameto install the extracted file at thedestination. The cached file is now either absent or complete — never truncated.
ResetCache:truefrom the vecindex roachtest (added in 2794f07 asa workaround for corrupt cache files). With atomic extraction the workaround is
no longer needed, and removing it eliminates the redundant 443MB download that
was triggering the race.
Fixes: #163471
Fixes: #159333
Test plan
vecindex/random-s/nodes=1/prefix=0locally with--local— passed (2250s).Release Justification: Test only fix.