matrixorigin / matrixorigin/matrixone
[Bug]: disk cache finalization UT flakes under CI scheduling load
- Dominant language
- Go
- Stars
- 1.9k
- Forks
- 311
- Avg merge
- 1d 3h
- Merged PRs (30d)
- 768
Description
### Is there an existing issue for the same bug?
- [x] I have checked the existing issues.
### Branch Name
main
### Commit ID
cde57d173b8928cc7979017e5fbf8fdb571c418d (PR #28782 run)
### Other Environment Information
- CI run: https://github.com/matrixorigin/matrixone/actions/runs/34693347823
- Job: Matrixone CI execution / UT Test on Ubuntu/x86
- The test ran in the race-sharded UT job. The same job later reported resource-heavy UT cgroup memory peaking near 16 GiB; no OOM kill was reported.
### Actual Behavior
`TestS3FSWriteDiskCacheFinalizeMode/async` failed at `pkg/fileservice/s3_fs_test.go:1111` with `disk-cache finalization did not start`. The test waited five seconds for the goroutine to reach the blocked file-sync barrier. The sibling `sync` subtest passed. This failure is unrelated to the Arrow LOAD test files changed by PR #28782.
### Expected Behavior
The test should preserve its async-versus-sync ordering assertions while remaining reliable when a race-enabled CI worker is heavily loaded. A scheduling delay should not be reported as a broken disk-cache transition unless the transition itself is stuck.
### Steps to Reproduce
1. Run the race-enabled UT suite on a busy Linux runner.
2. Execute `TestS3FSWriteDiskCacheFinalizeMode/async`.
3. Observe whether the goroutine reaches the injected file-sync barrier before the current five-second deadline.
### Additional information
This is one observed failure, so it does not yet prove a product-code defect or repeatable flake. The failure window was 13:17:53–13:17:59 UTC; the test hit its scheduling timeout before exercising the async/sync completion assertions. Please investigate the barrier/deadline behavior without removing those assertions.
Contributor guide
Assessment
This issue has not been assessed yet.