[
https://issues.apache.org/jira/browse/HDDS-3173?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17058743#comment-17058743
]
Arpit Agarwal commented on HDDS-3173:
-------------------------------------
Let's use the options you listed above, CMS GC while officially deprecated
still works on JDK 13, and is still recommended for applications that prefer
shorter pauses.
https://docs.oracle.com/en/java/javase/13/gctuning/available-collectors.html#GUID-C7B19628-27BA-4945-9004-EC0F08C76003
> Provide better default JVM options
> ----------------------------------
>
> Key: HDDS-3173
> URL: https://issues.apache.org/jira/browse/HDDS-3173
> Project: Hadoop Distributed Data Store
> Issue Type: Improvement
> Reporter: Marton Elek
> Priority: Major
>
> The GC pressure on Datanode is high because of the retry cache. I found
> crashes due to the long GC pauses. I started to use the following JVM
> parameters:
> {code}
> -server -XX:ParallelGCThreads=8 -XX:+UseConcMarkSweepGC
> -XX:CMSInitiatingOccupancyFraction=70 -XX:+UseCMSInitiatingOccupancyOnly
> {code}
> Which provide stable output.
> It would be great to detect the current version and add these parameters, if
> required.
> But there are two problems:
> * Different java versions support different flags
> * There could be conflicting flags (eg. if the user defines to use G1 we
> shouldn't add any other default parameters).
> I am not sure (yet) what is the best approach, I opened this issue to start a
> discussion.
> /cc [~arp]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: ozone-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: ozone-issues-h...@hadoop.apache.org
