[
https://issues.apache.org/jira/browse/SPARK-3304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14120030#comment-14120030
]
Kousuke Saruta commented on SPARK-3304:
---------------------------------------
Actually, I didn't see exception causing Reporter thread dead but unexpected
exception like OOM or any other exceptions due to communication error with RM
during resource negotiation caused by temporary traffic jam can be thrown, I
think.
> ApplicationMaster's Finish status is wrong when uncaught exception is thrown
> from ReporterThread
> ------------------------------------------------------------------------------------------------
>
> Key: SPARK-3304
> URL: https://issues.apache.org/jira/browse/SPARK-3304
> Project: Spark
> Issue Type: Bug
> Components: YARN
> Affects Versions: 1.1.0
> Reporter: Kousuke Saruta
>
> It's rare case but even though uncaught exception is thrown from Reporter
> thread, allocating containers, finish status is marked as SUCCEEDED.
> In addition, in YARN Cluster mode, if we don't notice Reporter thread is
> dead, we waits long time until timeout.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
