[
https://issues.apache.org/jira/browse/FLINK-20675?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17276128#comment-17276128
]
Piotr Nowojski commented on FLINK-20675:
----------------------------------------
FYI: this change caused UnalignedCheckpointITCase to become unstable.
https://issues.apache.org/jira/browse/FLINK-21215
Cancellation of Async part of the checkpoint is still not working correctly.
> Asynchronous checkpoint failure would not fail the job anymore
> --------------------------------------------------------------
>
> Key: FLINK-20675
> URL: https://issues.apache.org/jira/browse/FLINK-20675
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Checkpointing
> Affects Versions: 1.9.3, 1.10.2, 1.11.3, 1.12.0
> Reporter: Yun Tang
> Assignee: Yun Tang
> Priority: Critical
> Labels: pull-request-available
> Fix For: 1.11.4, 1.12.2, 1.13.0
>
>
> After FLINK-12364, no mater how many times of asynchronous part of checkpoint
> on task failed, the job itself would not fail by default:
> | Default behavior|Flink-1.5 —> Flink-1.8||Flink-1.9 -> Flink-1.12||
> |Synchronous part of checkpoint at task failed|Job failed|Job failed|
> |Asynchronous part of checkpoint at task failed| Job failed| Job would not
> fail|
> This error was because {{StreamTask}} use {{Exception}} instead of
> {{CheckpointException}} [when async part
> failed|https://github.com/apache/flink/blob/5125b1123dfcfff73b5070401dfccb162959080c/flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/StreamTask.java#L1118]
> as decline message. Thus checkpoint coordinator would call
> {{failPendingCheckpointDueToTaskFailure(pendingCheckpoint,
> CheckpointFailureReason.JOB_FAILURE, cause, executionAttemptID)}} to [process
> the declined
> checkpoint|https://github.com/apache/flink/blob/release-1.9/flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinator.java#L1316-L1323]:
> {code:java}
> if (cause == null) {
> failPendingCheckpointDueToTaskFailure(pendingCheckpoint,
> CheckpointFailureReason.CHECKPOINT_DECLINED, executionAttemptID);
> } else if (cause instanceof CheckpointException) {
> CheckpointException exception = (CheckpointException) cause;
> failPendingCheckpointDueToTaskFailure(pendingCheckpoint,
> exception.getCheckpointFailureReason(), cause, executionAttemptID);
> } else {
> failPendingCheckpointDueToTaskFailure(pendingCheckpoint,
> CheckpointFailureReason.JOB_FAILURE, cause, executionAttemptID);
> }
> {code}
> However, {{CheckpointFailureManager}} would [ignore the JOB_FAILURE
> reason|https://github.com/apache/flink/blob/5125b1123dfcfff73b5070401dfccb162959080c/flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointFailureManager.java#L108]
> and not count this failed checkpoint, which causes asynchronous checkpoint
> failure would not fail the job anymore.
>
> FLINK-16753 corrects the misleading message of JOB_FAILURE but the
> asynchronous checkpoint failure still cannot fail the job.
>
> As this bug exists too long, I decide to set it as critical instead of
> blocker level.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
