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 "yangcongqi (JIRA)" <ji...@apache.org> on 2015/12/17 10:23:46 UTC

[jira] [Commented] (YARN-2083) In fair scheduler, Queue should not been assigned more containers when its usedResource had reach the maxResource limit

    [ https://issues.apache.org/jira/browse/YARN-2083?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15061773#comment-15061773 ] 

yangcongqi commented on YARN-2083:
----------------------------------

There is a case: queue A maxResource is 1000mb, current usedResource is 900mb(less than maxResource), then a new request with 200mb came, fair scheduler would assign 200mb memory and usedResource=1100mb(over maxResource).

> In fair scheduler, Queue should not been assigned more containers when its usedResource had reach the maxResource limit
> -----------------------------------------------------------------------------------------------------------------------
>
>                 Key: YARN-2083
>                 URL: https://issues.apache.org/jira/browse/YARN-2083
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: fairscheduler
>    Affects Versions: 2.3.0
>            Reporter: Yi Tian
>              Labels: BB2015-05-TBR
>         Attachments: YARN-2083-1.patch, YARN-2083-2.patch, YARN-2083-3.patch, YARN-2083-4.patch, YARN-2083.patch
>
>
> In fair scheduler, FSParentQueue and FSLeafQueue do an assignContainerPreCheck to guaranty this queue is not over its limit.
> But the fitsIn function in Resource.java did not return false when the usedResource equals the maxResource.
> I think we should create a new Function "fitsInWithoutEqual" instead of "fitsIn" in this case.



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