[
https://issues.apache.org/jira/browse/YARN-3434?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14497065#comment-14497065
]
Wangda Tan commented on YARN-3434:
----------------------------------
bq. Are you suggesting we change the patch to modify ResourceLimits and pass
down rather then using the LimitsInfo class?
Yes, that's my suggested.
bq. at least not without adding the shouldContinue flag to it
Kind of, what I'm thinking is we can add "amountNeededUnreserve" to
ResourceLimits. canAssignToThisQueue/User will return boolean means
shouldContinue, and set "amountNeededUnreserve" (instead of limit, we don't
need to change limit). That very similar to your original logic and we don't
need the extra LimitsInfo. After we get the updated the ResourceLimit and pass
down, problem should be resolved.
Did I miss anything?
> 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)
