Class leakage on redeployment

[ngolovin]

Hello altogether,
I have memory leakage on redeployment of my own Seam application as well as on redeployment of the seam booking example -- the ClassLoader from the previous deployment doesn't get garbagecollected, preventing all the classes from being unloaded as well. That means that PermGen pool gets overflowed after several redeployments.

Has anybody had the same problem already?

[raja05]

Yeah, it throws a OOME -- PermGen Space Filled errors after some 10 - 15 redeployments. I had a read at the Resin site that it could be due to Debuggers if any , and I use a fair amount of debugging with IDE, so I thought that could lead to this, but yeah I get this error too.

[smokingapipe]

I experience exactly the same thing. I guess that's ok on my development box, but on my production box, I kill and restart JBoss, unless someone can suggest a way to avoid this leak.

Is there something that is preventing the classloader from being GC'ed? Or is this a problem with the JVM? Because it should be perfectly possible to GC a classloader.

[ssilvert]

"SmokingAPipe" wrote:
I experience exactly the same thing. I guess that's ok on my development box, but on my production box, I kill and restart JBoss, unless someone can suggest a way to avoid this leak.

Is there something that is preventing the classloader from being GC'ed? Or is this a problem with the JVM? Because it should be perfectly possible to GC a classloader.

It shouldn't leak in JBoss. If you have a step-by-step way to reproduce the problem, please open a Jira task and post a link to it here.

Thanks,

Stan

[marcioendo]

This is a known issue not directly related to SEAM but to EJB3:

http://jira.jboss.com/jira/browse/JBAS-3071