[
https://issues.apache.org/jira/browse/FLINK-14717?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16983367#comment-16983367
]
Chesnay Schepler commented on FLINK-14717:
------------------------------------------
How would you prioritize exceptions given the exception limit? I think you'd
end up with fairly inconsistent behavior from a users perspective.
Wouldn't it make more sense to add an vertex/attempt-based exception endpoint
(i.e., JobVertexExceptionsInfoHandler/SubtaskExecutionAttemptExceptionInfo)?
> JobExceptionsHandler show exceptions of prior attempts
> --------------------------------------------------------
>
> Key: FLINK-14717
> URL: https://issues.apache.org/jira/browse/FLINK-14717
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Web Frontend
> Reporter: lining
> Priority: Major
>
> *Current*
> The job's exceptions just show current attempt’s exceptions in web UI.(ps:
> [code|https://github.com/apache/flink/blob/34b5399f4effb679baabd8bca312cbf92ec34165/flink-runtime/src/main/java/org/apache/flink/runtime/rest/handler/job/JobExceptionsHandler.java#L97-L98])
> If the job failovers, we couldn't see any prior attempts' exceptions.
> *Proposal*
> We could use executionVertex.getPriorExecutionAttempt to get prior attempt in
> JobExceptionsHandler.
> {code:java}
> for (int i = task.getAttemptNumber() - 1; i >= 0; i--) {
> task = executionVertex.getPriorExecutionAttempt(i);
> }
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
