[
https://issues.apache.org/jira/browse/MAPREDUCE-4833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13535288#comment-13535288
]
Robert Parker commented on MAPREDUCE-4833:
------------------------------------------
Previously the Container did not send an event on kill if it was DONE, and
returned (essentially a no-op). This patch will send a TA_CONTAINER_CLEANED
event in all cases.
> Task can get stuck in FAIL_CONTAINER_CLEANUP
> --------------------------------------------
>
> Key: MAPREDUCE-4833
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-4833
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: applicationmaster, mrv2
> Affects Versions: 0.23.5
> Reporter: Robert Joseph Evans
> Assignee: Robert Parker
> Priority: Critical
> Attachments: MAPREDUCE4833-23.patch
>
>
> If an NM goes down and the AM still tries to launch a container on it the
> ContainerLauncherImpl can get stuck in an RPC timeout. At the same time the
> RM may notice that the NM has gone away and inform the AM of this, this
> triggers a TA_FAILMSG. If the TA_FAILMSG arrives at the TaskAttemptImpl
> before the TA_CONTAINER_LAUNCH_FAILED message then the task attempt will try
> to kill the container, but the ContainerLauncherImpl will not send back a
> TA_CONTAINER_CLEANED event causing the attempt to be stuck.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
