[jira] [Commented] (YARN-3508) Preemption processing occuring on the main RM dispatcher

["Wangda Tan (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/YARN-3508?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14573878#comment-14573878 ] 

Wangda Tan commented on YARN-3508:
----------------------------------

Trying to better understand this problem: I'm not sure where is bottleneck. If CapacityScheduler becomes bottleneck, move preemption events out of main RM dispatcher doesn't help. This approach only helps when main dispatcher is bottleneck.

And a parallel thing we can do is to optimize number of preemption event. Currently, if a container sits in to-preempt list, before it is get preempted, one event will be sent to scheduler for every few seconds, we can reduce frequency of this event.

> Preemption processing occuring on the main RM dispatcher
> --------------------------------------------------------
>
>                 Key: YARN-3508
>                 URL: https://issues.apache.org/jira/browse/YARN-3508
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: resourcemanager, scheduler
>    Affects Versions: 2.6.0
>            Reporter: Jason Lowe
>            Assignee: Varun Saxena
>         Attachments: YARN-3508.002.patch, YARN-3508.01.patch
>
>
> We recently saw the RM for a large cluster lag far behind on the AsyncDispacher event queue.  The AsyncDispatcher thread was consistently blocked on the highly-contended CapacityScheduler lock trying to dispatch preemption-related events for RMContainerPreemptEventDispatcher.  Preemption processing should occur on the scheduler event dispatcher thread or a separate thread to avoid delaying the processing of other events in the primary dispatcher queue.



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