[ https://issues.apache.org/jira/browse/SPARK-28239?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Deegue updated SPARK-28239:
---------------------------
Summary: Allow TCP connections created by shuffle service auto close on YARN NodeManagers
(was: Make TCP connections created by shuffle service auto close on YARN NodeManagers)
> Allow TCP connections created by shuffle service auto close on YARN NodeManagers
> --------------------------------------------------------------------------------
>
> Key: SPARK-28239
> URL: https://issues.apache.org/jira/browse/SPARK-28239
> Project: Spark
> Issue Type: Improvement
> Components: Shuffle, YARN
> Affects Versions: 2.4.0
> Environment: Hadoop2.6.0-CDH5.8.3(netty3)
> Spark2.4.0(netty4)
> Configs:
> spark.shuffle.service.enabled=true
> Reporter: Deegue
> Priority: Minor
> Attachments: screenshot-1.png, screenshot-2.png
>
>
> When executing shuffle tasks, TCP connections(on port 7337 by default) will be established
by shuffle service.
> It will like:
> !screenshot-1.png!
> However, some of the TCP connections are still busy when the task is actually finished.
These connections won't close automatically until we restart the NodeManager process.
> Connections pile up and NodeManagers are getting slower and slower.
> !screenshot-2.png!
> These unclosed TCP connections stay busy and it seem doesn't take effect when I set ChannelOption.SO_KEEPALIVE
to true according to [SPARK-23182|https://github.com/apache/spark/pull/20512].
> So the solution is setting ChannelOption.AUTO_CLOSE to true, and after which our cluster(running
10000+ jobs / day) is processing normally.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org
|