Memory Leak in Wildfly 8.2 on remote EJB access from Tomcat 7

[maderhan]

I found this issue as fixed:

[REM3-200] Memory leak in org.xnio.ByteBufferSlicePool - JBoss Issue Tracker

I patched from wf 8.1 to 8.2 and it seems to be that the memory leak still remains!
A large amount of byte[] data could not be garbage collected.

Only the described workaround "-Djboss.remoting.pooled-buffers=false" solves the memory leak.

The remoting module 4.0.6 is used for sure.

Our setup:
Wildfly 8.2 (java 8 u 20 with 4G MEM) on redhat linux
Tomcat 7 (java 7 u 55 with 3 G MEM) on redhat linux

Our scenario: Tomcat application connects to ejb deployed on Wildfly via scoped EJB Client context

The result: Tomcat runs OOM and Wildfly runs OOE during one day.

Hans.

[jaikiran]

If it OOMs on Tomcat, then you have to update the (remoting) libraries on Tomcat too.

[maderhan]

I took the jboss-client.jar from the 8.2 release for sure .

As I mentioned before, only the workaround solves the problem - on both sides (wildfly and tomcat).

[maderhan]

Just for information:

I found out that the original wf8.2 download has the jboss-ejb-client-2.0.1.Final.jar module included.

What about this version jboss-ejb-client-2.0.2.Final.jar:

http://search.maven.org/#artifactdetails%7Corg.jboss%7Cjboss-ejb-client%7C2.0.2.Final%7Cjar

At the moment I could run the EJB driven applications for one week.

I will apply the jboss-ejb-client-2.0.2.Final.jar and see what will happen.

[basmaabd]

Hello,

I have a similar setup as yours and I am running into the same issue. The workaround "-Djboss.remoting.pooled-buffers=false" makes it better and tomcat runs longer before giving an OOM. But when I  run a load test, tomcat crashes due to an OOM. This did not happen with jboss version 6. Did changing the jar to  jboss-ejb-client-2.0.2.Final.jar  help you ? Or, did you find any other solution?

Thanks.