[
https://issues.apache.org/jira/browse/YARN-6112?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15840585#comment-15840585
]
Yufei Gu commented on YARN-6112:
--------------------------------
There is an another duration measurement outside of the function update()
(shows in following code), which did the measurement outside the writelock.
It's a bit surprise me, by looking at the original design in YARN-2352. There
are two because there is one more line after {{update()}}. We should remove one
of measurements since they are identical.
{code}
long start = getClock().getTime();
update();
long duration = getClock().getTime() - start;
fsOpDurations.addUpdateThreadRunDuration(duration);
{code}
> fsOpDurations.addUpdateCallDuration() should be independent to LOG level
> ------------------------------------------------------------------------
>
> Key: YARN-6112
> URL: https://issues.apache.org/jira/browse/YARN-6112
> Project: Hadoop YARN
> Issue Type: Bug
> Components: fairscheduler
> Reporter: Yufei Gu
> Assignee: Yufei Gu
> Attachments: YARN-6112.001.patch
>
>
> In the update thread of Fair Scheduler, the
> {{fsOpDurations.addUpdateCallDuration()}} records the duration of
> {{update()}}, it should be independent to LOG level. YARN-4752 put the it
> inside a {{LOG.isDebugEnabled()}} block. Not sure any particular reason to do
> that. cc [~kasha]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
