[ 
https://issues.apache.org/jira/browse/YARN-3434?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14487416#comment-14487416
 ] 

Thomas Graves commented on YARN-3434:
-------------------------------------

[~wangda]  I'm not sure I follow what are saying?  The reservations are already 
counted in the users usage and we do consider reserved when doing the user 
limit calculations.   Look at LeafQueue.assignContainers call to 
allocateResource is where it ends up adding to user usage.    The 
canAssignToUser is where it does user limit check and substracts the 
reservations off to see if it can continue.  

Note I do think we should just get rid of the config for 
reservationsContinueLooking, but that is a separate issue.

> Interaction between reservations and userlimit can result in significant ULF 
> violation
> --------------------------------------------------------------------------------------
>
>                 Key: YARN-3434
>                 URL: https://issues.apache.org/jira/browse/YARN-3434
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: capacityscheduler
>    Affects Versions: 2.6.0
>            Reporter: Thomas Graves
>            Assignee: Thomas Graves
>         Attachments: YARN-3434.patch
>
>
> ULF was set to 1.0
> User was able to consume 1.4X queue capacity.
> It looks like when this application launched, it reserved about 1000 
> containers, each 8G each, within about 5 seconds. I think this allowed the 
> logic in assignToUser() to allow the userlimit to be surpassed.



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

Reply via email to