[
https://issues.apache.org/jira/browse/YARN-3278?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14340711#comment-14340711
]
Wangda Tan commented on YARN-3278:
----------------------------------
One possible solution is forced update current-max-limit of *all* queues when
resource updated (container allocated/released, etc.). But this may lead to
performance regression, we should verify this plan is efficiency before doing
that.
> Queue's current-max-limit should be updated when container allocated/released
> in another queue
> ----------------------------------------------------------------------------------------------
>
> Key: YARN-3278
> URL: https://issues.apache.org/jira/browse/YARN-3278
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: capacityscheduler, resourcemanager
> Reporter: Wangda Tan
> Assignee: Wangda Tan
>
> This is a problem after YARN-2008, queue's current-max-limit (it was called
> maxAvailCapacity in YARN-2008) will be updated when queue is traversed by
> scheduler, so application can get headroom based on staled current-max-limit.
> For example:
> {code}
> r
> / \
> a b
> / \
> a1 b1
> ^ ^
> | |
> app1 app2
> {code}
> We have above queue structure, two apps in the cluster: app1 in a1, and app2
> in b1. If scheduler always travers b1 only, and continously allocate
> containers in b1, app1's headroom will not get updated.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
