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 "Jason Lowe (JIRA)" <ji...@apache.org> on 2015/05/12 15:22:01 UTC

[jira] [Commented] (YARN-3631) Tracking URI for killed applications

    [ https://issues.apache.org/jira/browse/YARN-3631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14539816#comment-14539816 ] 

Jason Lowe commented on YARN-3631:
----------------------------------

This is strongly related to, if not a duplicate of, MAPREDUCE-5502 and/or YARN-2261.  The problem is that YARN provides no ability for an application to perform cleanup processing outside of what the ApplicationMaster can perform.  The MapReduce framework relies on the AM to move the .jhist file to the intermediate directory for the JHS to pick up for the job's history.  If the application is killed before the AM runs or the last app attempt crashes or is otherwise incapacitated (e.g. expires from the RM's point of view) then there will be no history for the job on the JHS.


> Tracking URI for killed applications
> ------------------------------------
>
>                 Key: YARN-3631
>                 URL: https://issues.apache.org/jira/browse/YARN-3631
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Anushri
>            Priority: Minor
>
> If some running application is killed we can't track it in JHS
> but still at UI the tracking URI even for killed application is given as "History" 
> Also the REST API returns the link to trace it in "cluster- applications" list.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)