[
https://issues.apache.org/jira/browse/CASSANDRA-18841?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17764929#comment-17764929
]
Doug Rohrer commented on CASSANDRA-18841:
-----------------------------------------
Yeah - not sure why the runsOnInstance is there in the patch - it's really just
wrapping the call in {{sync}} - I'll update it to make sure it's clean in the
morning.
Some more context as to why this fixed the problem we saw, there was some new
code added in the above-mentioned commit that eventually creates a new instance
of a {{o.netty.util.internal.InternalThreadLocalMap}} with the instance class
loader, but in the main thread. By calling {{sync}} we execute the code on the
instance's executor thread so when the instance is shut down the
{{InternalThreadLocalMap}} is removed appropriately.
> InstanceClassLoader leak in 5.0/trunk
> -------------------------------------
>
> Key: CASSANDRA-18841
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18841
> Project: Cassandra
> Issue Type: Bug
> Components: Test/dtest/java
> Reporter: Doug Rohrer
> Assignee: Doug Rohrer
> Priority: Normal
> Fix For: 5.0.x, 5.x
>
> Attachments: trunk_ThreadLocal_leak.patch
>
>
> Something in the 5.0/trunk branches has caused an in-jvm dtest
> InstanceClassLoader leak - it appears to have something to do with the Mutual
> TLS Authenticator (f078c02cb58bddd735490b07548f7352f0eb09aa) but nothing in
> that commit, so far, has stood out as causing issues.
> The culprit class appears to be
> {{io.netty.util.internal.InternalThreadLocalMap}}, which seems to no be
> removed when the threads stops for some reason.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
