[
https://issues.apache.org/jira/browse/YARN-11401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17649255#comment-17649255
]
ASF GitHub Bot commented on YARN-11401:
---------------------------------------
Daniel-009497 opened a new pull request, #5241:
URL: https://github.com/apache/hadoop/pull/5241
Currently, there is only one thread pool to handle AM launch and cleanup
event in ResourceManager.
We found an issue in our cluster, in some cases, too many cleanup events
will block AM launcher event and make it stuck for a long time.
So in this PR, we divide the shared thread pool into two separated ones to
handle AM event of different kinds in case that a flood events of cleanup
events blocking launcher event for a long time and vice versa, which will badly
decay throughput of ResourceManager.
> Separate AppMaster cleanup events and launcher event into different resource
> pools
> ----------------------------------------------------------------------------------
>
> Key: YARN-11401
> URL: https://issues.apache.org/jira/browse/YARN-11401
> Project: Hadoop YARN
> Issue Type: Improvement
> Reporter: Daniel Ma
> Priority: Major
>
> Currently, there is only one thread pool to handle AM launch and cleanup
> event by ResourceManager,
> In some cases, too many cleanup event will lead to AM launch stuck for a long
> time.
> So in this patch, We divide the shared thread pool into two separated ones to
> handle different event in case that a flood event of one kind may block other
> kinds for a long time.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
