[jira] [Commented] (FLINK-3050) Add custom Exception type to suppress job restarts

["Chesnay Schepler (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-3050?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15021054#comment-15021054 ] 

Chesnay Schepler commented on FLINK-3050:
-----------------------------------------

Would it be sufficient to modify this bit:
{code:title=ExecutionGraph.jobVertexInFinalState() - Line 918}
else if (current == JobStatus.FAILING) {
	if (numberOfRetriesLeft > 0 && !(getFailureCause() instanceof UnrecoverableFailure) && transitionState(current, JobStatus.RESTARTING)) {
		numberOfRetriesLeft--;
{code}

> Add custom Exception type to suppress job restarts
> --------------------------------------------------
>
>                 Key: FLINK-3050
>                 URL: https://issues.apache.org/jira/browse/FLINK-3050
>             Project: Flink
>          Issue Type: Improvement
>          Components: Distributed Runtime
>    Affects Versions: 0.10.0
>            Reporter: Ufuk Celebi
>            Priority: Minor
>             Fix For: 1.0.0
>
>
> In case of failures and configured execution retries, the job will be be restarted even in cases when the failure is not recoverable.
> We can add a custom Exception type like UnrecoverableFailure in order to suppress restarts in certain cases. The execution graph restart logic can check the failure type on recovery and skip the restarting.
> This Exception can be used both by the system and the user.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)