[
https://issues.apache.org/jira/browse/MAPREDUCE-5848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13987061#comment-13987061
]
Jason Lowe edited comment on MAPREDUCE-5848 at 5/1/14 10:11 PM:
----------------------------------------------------------------
Sure, I think that's a reasonable approach. It's definitely an incremental
improvement from what we have today.
The kill event should have a useful diagnostic message attached to it so users
can tell that it was preempted, i.e.: instead of a raw TA_KILL event it should
send a TaskAttemptKillEvent with a diagnostic. Otherwise I think users will be
confused as to why the task was killed (e.g.: was it preempted vs. aborted?)
and may have to dig through a rather large AM log to sort that out.
It'd also be nice to add a unit test in TestRMContainerAllocator.
was (Author: jlowe):
Sure, I think that's a reasonable approach. It's definitely an incremental
improvement from what we have today.
The kill event have a useful diagnostic message attached to it so users can
tell that it was preempted, i.e.: instead of a raw TA_KILL event it should send
a TaskAttemptKillEvent with a diagnostic. Otherwise I think users will be
confused as to why the task was killed (e.g.: was it preempted vs. aborted?)
and may have to dig through a rather large AM log to sort that out.
It'd also be nice to add a unit test in TestRMContainerAllocator.
> MapReduce counts forcibly preempted containers as FAILED
> --------------------------------------------------------
>
> Key: MAPREDUCE-5848
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5848
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Affects Versions: 2.1.0-beta
> Reporter: Carlo Curino
> Assignee: Subramaniam Krishnan
> Attachments: YARN-1958.patch
>
>
> The MapReduce AM is considering a forcibly preempted container as FAILED,
> while I think it should be considered as KILLED (i.e., not count against the
> maximum number of failures).
--
This message was sent by Atlassian JIRA
(v6.2#6252)
