[
https://issues.apache.org/jira/browse/YARN-4969?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15267765#comment-15267765
]
Wangda Tan commented on YARN-4969:
----------------------------------
[~vvasudev],
I would prefer not to do that, since currently scheduler depends on other
component, for example, ApplicationMasterService. If we move scheduler logs to
a separate file, it gonna be inconvenient to view context of logs. For example,
when application submit to RM and when application is submit to scheduler. We
need to grep two log files to find answers like this.
Thoughts?
> Fix more loggings in CapacityScheduler
> --------------------------------------
>
> Key: YARN-4969
> URL: https://issues.apache.org/jira/browse/YARN-4969
> Project: Hadoop YARN
> Issue Type: Bug
> Components: capacity scheduler
> Reporter: Wangda Tan
> Assignee: Wangda Tan
> Attachments: YARN-4969.1.patch
>
>
> YARN-3966 did logging cleanup for Capacity Scheduler before, however,
> there're some loggings we need to improvement:
> Container allocation / complete / reservation / un-reserve messages for every
> hierarchy (app/leaf/parent-queue) should be printed at INFO level:
> I'm debugging one issue that root queue's resource usage could be negative,
> it is very hard to reproduce, so we cannot enable debug logging since RM
> start, size of log cannot be fit in a single disk.
> Existing CS prints INFO message when container cannot be allocated, such as
> re-reservation / node heartbeat, etc. we should avoid printing such message
> at INFO level.
--
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
