[
https://issues.apache.org/jira/browse/YARN-7397?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16221247#comment-16221247
]
Yufei Gu edited comment on YARN-7397 at 10/26/17 11:23 PM:
-----------------------------------------------------------
Nice finding! Thanks for working on this. The patch looks good to me. I was
wandering {{getAppWeight()}} isn't necessary to be in class FairScheduler,
{{FSAppAttempt}} seems a good place to hold that, but it may involve some lock
rewriting. I don't think it is worth to do in this jira if it is too
complicated. +1.
was (Author: yufeigu):
Nice finding! Thanks for working on this. The patch looks good to me. I was
wandering {{getAppWeight()}} isn't necessary to be in class FairScheduler,
{{FSAppAttempt}} seems a good place to hold that, but it may involve some lock
rewriting. I don't think it is worth to do in this jira if it is two
complicated.
> Reduce lock contention in FairScheduler#getAppWeight()
> ------------------------------------------------------
>
> Key: YARN-7397
> URL: https://issues.apache.org/jira/browse/YARN-7397
> Project: Hadoop YARN
> Issue Type: Improvement
> Components: fairscheduler
> Affects Versions: 3.0.0-beta1
> Reporter: Daniel Templeton
> Assignee: Daniel Templeton
> Attachments: YARN-7397.001.patch
>
>
> In profiling the fair scheduler, a large amount of time is spent waiting to
> get the lock in {{FairScheduler.getAppWeight()}}, when the lock isn't
> actually needed. This patch reduces the scope of the lock to eliminate that
> contention.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
