MartijnVisser opened a new pull request, #28653:
URL: https://github.com/apache/flink/pull/28653
## What is the purpose of the change
`SinkV2ITCase.writerAndCommitterExecuteInStreamingModeWithScaling` is the
dominant flaky failure on the release-1.20 nightlies (19 of the last 25 failed
runs, 37 test errors, across all JDK/profile legs). The test body always
passes; the failure is a teardown race. The checkpoint directory is injected as
a per-method JUnit `@TempDir`, but the class-shared MiniCluster keeps
discarding checkpoint state asynchronously after `requestJobResult().get()`
returns (on a successful FINISH the job's completed checkpoints are discarded
via the CheckpointCoordinator/CheckpointsCleaner IO executor as fire-and-forget
work). JUnit's temp-directory deletion then races with Flink's discard and
fails with `IOException: Failed to delete temp directory` (suppressed
`NoSuchFileException`, i.e. files vanishing mid-walk).
The race is latent on master too (identical `@TempDir` wiring and
`requestJobResult().get()`); it is fixed here first so the change can be
backported to the release branches where it currently manifests (FLINK-40074).
## Brief change log
- Manage the scaling test's checkpoint directory explicitly
(`Files.createTempDirectory`) instead of injecting it via `@TempDir`.
- Clean it up in a `finally` block with `FileUtils.deleteDirectoryQuietly`,
which is safe against concurrent deletions and never lets a cleanup failure
mask a real assertion failure.
- No change to the test's assertions or job logic.
## Verifying this change
This change is already covered by the existing test
`SinkV2ITCase.writerAndCommitterExecuteInStreamingModeWithScaling`. Verified
locally: the method passes (3/3 parameterized cases) and `spotless:check` is
clean. On release-1.20, where the race manifests, the backported change ran
green across repeated executions.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): no
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: no
- The serializers: no
- The runtime per-record code paths (performance sensitive): no
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Kubernetes/Yarn, ZooKeeper: no (test-only change)
- The S3 file system connector: no
## Documentation
- Does this pull request introduce a new feature? no
- If yes, how is the feature documented? not applicable
---
##### Was generative AI tooling used to co-author this PR?
- [X] Yes (Claude Fable 5)
Generated-by: Claude Fable 5
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
