Efrat19 opened a new pull request, #29517:
URL: https://github.com/apache/flink/pull/29517
The test was originally meant to enable checkpointing with a large interval,
so no checkpoint is effectively triggered.
Since FLINK-38939 the coordinator re-schedules the periodic trigger based on
mean pause. With the default min pause of 0 a periodic checkpoint fires as soon
as all tasks are running.
That early checkpoint lets the subtasks that were finished on restore
(sourceA, passA -> sinkA) complete and exit immediately. The end-of-data
checkpoint is queued behind it and triggered against a plan that still lists
those subtasks as RUNNING. If one of them reaches FINISHED on the TaskManager
before the trigger arrives, it declines with
CHECKPOINT_DECLINED_TASK_NOT_READY. That reason is ignored by the
CheckpointFailureManager, the end-of-data trigger is not retried, and no
periodic checkpoint follows, so sourceB waits forever in StreamTask.afterInvoke
for a final checkpoint and the passB -> sinkB subtasks wait for its
end-of-partition.
Set the min pause of the restored job to one year, the maximum the
coordinator accepts. The pause gates periodic triggers only, so the end-of-data
checkpoint still fires at once, while every scheduler start or restart now
waits longer than the test can run. All tasks then wait for the same final
checkpoint and none can finish before it is taken, which removes the race
instead of recovering from it.
Generated-by: Claude Code (Claude Fable 5.1)
## What is the purpose of the change
Fix flakiness in `testRestoreAfterSomeTasksFinished` caused by periodic
checkpointing
## Brief change log
- Effectively re-disable periodic checkpointing for the test by increasing
`minPauseBetweenCheckpoints`
## Verifying this change
This change is a trivial rework / code cleanup without any test coverage.
## 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
- 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 (please specify the tool below)
Co-Generated-by: Claude Code (Claude Fable 5.1)
--
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]
