[jira] [Commented] (YARN-8328) NonAggregatingLogHandler needlessly waits upon shutdown if delayed deletion is scheduled

["Jonathan Eagles (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/YARN-8328?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16481478#comment-16481478 ] 

Jonathan Eagles commented on YARN-8328:
---------------------------------------

Attaching a patch that is to be used as discussion and illustrate what is going on. I'm not assuming this is the correct approach since the could be many ways to address this issue.

> NonAggregatingLogHandler needlessly waits upon shutdown if delayed deletion is scheduled
> ----------------------------------------------------------------------------------------
>
>                 Key: YARN-8328
>                 URL: https://issues.apache.org/jira/browse/YARN-8328
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Jonathan Eagles
>            Priority: Major
>         Attachments: YARN-8328.001.patch
>
>
> This happens frequently in the MiniYarnCluster setup where a job completes and then the cluster is shutdown. Often the jobs are scheduled to delay deletion of the log files so that container logs are available for debugging. A scheduled log deletion has a default value of 3 hours. When the NonAggregating LogHandler is stopped it waits 10 seconds for all delayed scheduled log deletions to occur and then shuts down.
> A test developer has to make a trade of to either 1) set the log deletion to 0 to have the tests quickly shutdown but get no container logs to debug or 2) keep the log deletion at 3 hours and incur the 10 second timeout for each test suit.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org