[ 
https://issues.apache.org/jira/browse/YARN-6246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15947815#comment-15947815
 ] 

Daniel Templeton commented on YARN-6246:
----------------------------------------

Thanks, [~kasha].  Looking at the patch, I'm concerned about two elements of 
state that are no longer protected by a lock: {{FSAppAttempt.demand}} and 
{{FSLeafQueue.lastTimeAtFairShare}}.  It looks to me like both are used from 
different threads with no locking.  {{FSLeafQueue.dumpStateInternal()}} appears 
to be the primary problem.

> Identifying starved apps does not need the scheduler writelock
> --------------------------------------------------------------
>
>                 Key: YARN-6246
>                 URL: https://issues.apache.org/jira/browse/YARN-6246
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: fairscheduler
>    Affects Versions: 2.9.0
>            Reporter: Karthik Kambatla
>            Assignee: Karthik Kambatla
>         Attachments: YARN-6246.001.patch, YARN-6246.002.patch, 
> YARN-6246.003.patch
>
>
> Currently, the starvation checks are done holding the scheduler writelock. We 
> are probably better of doing this outside. 



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to