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 "Sunil G (JIRA)" <ji...@apache.org> on 2017/05/01 12:41:04 UTC

[jira] [Commented] (YARN-2113) Add cross-user preemption within CapacityScheduler's leaf-queue

    [ https://issues.apache.org/jira/browse/YARN-2113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15990757#comment-15990757 ] 

Sunil G commented on YARN-2113:
-------------------------------

Adding further to [~leftnoteasy]'s points.

In patch 11/12, we tried to identify container which could bring user's usage under UL.
Something like below
 {code}
if ((user.getUsed() - container.getAllocatedResource()) < user.getUserLimit()) {
   // Skip container
}
{code}. This code was added in container selection logic. Hence same code was hitting for priority/user-limit preemption as well. So if there is only one or 2 user's with  100% as MULP, UL will be same or more than queue capacity. 

Now if we have a high priority app within same queue (same user also), due to the added check like above, preemption selection will fail. Hence I added a logic that if user.getUsed() is already under User Limit, then we can preempt. (Because this must be due to priority based preemption).

I think we can chat offline to be in sync about this case faster and can post the summary here.

> Add cross-user preemption within CapacityScheduler's leaf-queue
> ---------------------------------------------------------------
>
>                 Key: YARN-2113
>                 URL: https://issues.apache.org/jira/browse/YARN-2113
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: scheduler
>            Reporter: Vinod Kumar Vavilapalli
>            Assignee: Sunil G
>         Attachments: IntraQueue Preemption-Impact Analysis.pdf, TestNoIntraQueuePreemptionIfBelowUserLimitAndDifferentPrioritiesWithExtraUsers.txt, YARN-2113.0001.patch, YARN-2113.0002.patch, YARN-2113.0003.patch, YARN-2113.0004.patch, YARN-2113.0005.patch, YARN-2113.0006.patch, YARN-2113.0007.patch, YARN-2113.0008.patch, YARN-2113.0009.patch, YARN-2113.0010.patch, YARN-2113.0011.patch, YARN-2113.0012.patch, YARN-2113.0013.patch, YARN-2113.apply.onto.0012.ericp.patch, YARN-2113.v0.patch
>
>
> Preemption today only works across queues and moves around resources across queues per demand and usage. We should also have user-level preemption within a queue, to balance capacity across users in a predictable manner.



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

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