[jboss-jira] [JBoss JIRA] (REMJMX-94) Memory leak in remoting-jmx

[Darran Lofthouse (JIRA)]

    [ https://issues.jboss.org/browse/REMJMX-94?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13056659#comment-13056659 ] 

Darran Lofthouse commented on REMJMX-94:
----------------------------------------

[~dmlloyd] Might need to run this one by you - debugging this myself with the XNIO version currently used in WildFly (And Core) I am still seeing the same behaviour that Brad describes with the classloaders being kept around with references leading back to ThreadLocalCache of ByteBufferSlicePool.

Within Remoting JMX I have double checked and there are no remaining places that I fail to clean up resources that I created so not sure if there is anything else I can do to clean this up myself.  

I do have a heap dump after modifying the test case if you want.


> Memory leak in remoting-jmx
> ---------------------------
>
>                 Key: REMJMX-94
>                 URL: https://issues.jboss.org/browse/REMJMX-94
>             Project: Remoting JMX
>          Issue Type: Bug
>    Affects Versions: 1.1.3.Final
>            Reporter: Libor Zoubek
>            Assignee: Darran Lofthouse
>              Labels: memory_leak
>             Fix For: 2.0.1.CR2
>
>
> In RHQ plugin We're experiencing a memory leak when we talk to EAP over remoting-jmx. Leak is very notable once we restart our plugin container (each plugin has in it's own classloader) Classes owned by RHQ Server plugin don't get GCed.
> I am able to reproduce mem leak without RHQ agent and plugin infrastructure repeating following steps:
> 1. start new thread
> 2. assign URL classloader to this thread (point it to jboss-client.jar)
> 3. connect & disconnect to jmx-remoting endpoint
> In this test no classes don't get GCed which after serveral iterations ends up in OOM.
> My test/example is here https://github.com/lzoubek/remotingjmx-leak/blob/master/src/main/java/org/jboss/remotingjmx/memleak/test/Main.java
> I am not sure if memleak is in remoting-jmx or some other underlying stuff. 
> I thought https://issues.jboss.org/browse/REM3-200 was related, but workaround mentioned does not help.



--
This message was sent by Atlassian JIRA
(v6.3.11#6341)