You are viewing a plain text version of this content. The canonical link for it is here.

Posted to yarn-issues@hadoop.apache.org by "Bikas Saha (JIRA)" <ji...@apache.org> on 2013/04/02 22:21:15 UTC

[jira] [Commented] (YARN-529) Succeeded MR job is retried by RM if finishApplicationMaster() call fails

    [ https://issues.apache.org/jira/browse/YARN-529?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13620235#comment-13620235 ] 

Bikas Saha commented on YARN-529:
---------------------------------

This problem is related to RM Restart but independent of it. Even without restart, if for some reason, during MR app master shutdown, if unregister from RM fails, then the app master will continue and delete staging dir etc. Since RM did not get an unregister, it will retry the MR app and all subsequent attempts will fail.
                
> Succeeded MR job is retried by RM if finishApplicationMaster() call fails
> -------------------------------------------------------------------------
>
>                 Key: YARN-529
>                 URL: https://issues.apache.org/jira/browse/YARN-529
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>            Reporter: Jian He
>            Assignee: Jian He
>
> MR app master will clean staging dir, if the job is already succeeded and asked to reboot. If the finishApplicationMaster call fails, RM will consider this job unfinished and launch further attempts, further attempts will fail because staging dir is cleaned

--
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