[
https://issues.apache.org/jira/browse/MAPREDUCE-2783?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13122066#comment-13122066
]
Eric Payne commented on MAPREDUCE-2783:
---------------------------------------
It looks like we are talking only of the rare case where the AppMaster dies
somehow, right? For failed jobs, the 'Tracking UI' column looks like it is set
correctly to point to the job history page for that job.
In the case when a job fails, it is the AM that sends the unregister event to
the RM telling the RM to change the tracking URL. However, in the use case we
are addressing, the AM has died. I've looked into alterative ways to get the
job history URL for a job in this case, but I think it would involve having
other daemons try to recreate the AM's unregister event.
To me, since this is a narrow use case, I think it is sufficient to just
"null-out" the tracking URL, which will cause the scheduler UI to put
UNASSIGNED in the 'Tracking UI' column, which will not be a link.
> mr279 job history handling after killing application
> ----------------------------------------------------
>
> Key: MAPREDUCE-2783
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2783
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mrv2
> Reporter: Thomas Graves
> Assignee: Eric Payne
> Priority: Critical
>
> The job history/application tracking url handling during kill is not
> consistent. Currently if you kill a job that was running the tracking url
> points to job history, but job history server doesn't have the job.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
