You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Qi Zhu (Jira)" <ji...@apache.org> on 2021/04/03 16:25:00 UTC
[jira] [Comment Edited] (YARN-9927) RM multi-thread event
processing mechanism
[ https://issues.apache.org/jira/browse/YARN-9927?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17314294#comment-17314294 ]
Qi Zhu edited comment on YARN-9927 at 4/3/21, 4:24 PM:
-------------------------------------------------------
[~gandras] [~pbacsko]
Updated a patch, each eventType will have single async dispatcher thread. I will add unit test later.
I think we first realize the multi thread dispatcher for each eventType, so that the event queue will not be a problem. And about the thread number, i counted the eventType number, it is in reasonable range.
Then we can move to that, how can we consume the event queue faster in another jira.
What's your opinion? :D
Thanks.
was (Author: zhuqi):
[~gandras] [~pbacsko]
Updated a patch, each eventType will have single async dispatcher thread. I will add unit test later.
I think we first realize the multi thread dispatcher for each eventType, so that the event queue will not be a problem. And about the thread number, i counted the eventType number, it is in reasonable range.
Then we can move to that, how can we consume the event queue faster.
> RM multi-thread event processing mechanism
> ------------------------------------------
>
> Key: YARN-9927
> URL: https://issues.apache.org/jira/browse/YARN-9927
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: yarn
> Affects Versions: 3.0.0, 2.9.2
> Reporter: hcarrot
> Assignee: Qi Zhu
> Priority: Major
> Attachments: RM multi-thread event processing mechanism.pdf, YARN-9927.001.patch, YARN-9927.002.patch
>
>
> Recently, we have observed serious event blocking in RM event dispatcher queue. After analysis of RM event monitoring data and RM event processing logic, we found that
> 1) environment: a cluster with thousands of nodes
> 2) RMNodeStatusEvent dominates 90% time consumption of RM event scheduler
> 3) Meanwhile, RM event processing is in a single-thread mode, and It results in the low headroom of RM event scheduler, thus performance of RM.
> So we proposed a RM multi-thread event processing mechanism to improve RM performance.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org