[
https://issues.apache.org/jira/browse/CASSANDRA-19976?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17886573#comment-17886573
]
Michael Semb Wever commented on CASSANDRA-19976:
------------------------------------------------
Has anyone any experience running with this flag (in production) ?
> Consider setting -Djdk.nio.MaxCachedBufferSize by default
> ---------------------------------------------------------
>
> Key: CASSANDRA-19976
> URL: https://issues.apache.org/jira/browse/CASSANDRA-19976
> Project: Cassandra
> Issue Type: Improvement
> Components: Local/Config
> Reporter: Dmitry Konstantinov
> Priority: Normal
>
> Long time ago CASSANDRA-8670 added a fix of native memory leaking through
> JDK-level thread-local caches for direct buffers.
> From one side since that time there were a lot of changes in Cassandra IO and
> networking logic. From another side
> https://bugs.openjdk.org/browse/JDK-8147468 - a JDK-level fix for the issue
> (by adding of -Djdk.nio.MaxCachedBufferSize option) was implemented some time
> after CASSANDRA-8670 and backported to JDK8.
> Since that time there were many places where the JDK option has been
> mentioned in a similar context as a possible way to protect against direct
> byte buffer memory leakage:
> - CASSANDRA-10689
> - CASSANDRA-13931
> - CASSANDRA-15006
> - CASSANDRA-15400
> -
> https://support.datastax.com/s/article/JVM-OOM-direct-buffer-errors-affected-by-unlimited-javanio-cache
> (DSE currently sets it by default to 1Mb)
> -
> https://github.com/k8ssandra/cass-operator/blob/19ed28196f2549a763ec7d3794328565f6d4b127/docs/user/jvm_server_configuration.md?plain=1#L207
> I have created the ticket to consider if it makes sense to add
> -Djdk.nio.MaxCachedBufferSize by default to Cassandra jvm-server.options.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
