4 Replies Latest reply on Jan 20, 2015 10:24 AM by Johannes Hoffmeister

    OutOfMemory with Infinispan 5.3.0 in glassfish v3 cluster

    InfinispanUser 0815 Newbie

      There was an memory leak in AbstractJBossMarshaller (see issue https://issues.jboss.org/browse/ISPN-2372) which should already been fixed with version 5.2.0 Beta2 of Infinispan.

       

      Currently we are using Infinispan Version 5.3.0 final, but still we are running out of memory with our application. We have a Java Enterprise Archive Application (EAR) deployed on a glassfish v3 (Oracle GlassFish Server 3.1.2.6 (build 3)) cluster with 2 servers.

      The JDK is jdk1.7.0_25_iaik_5.1.

       

      The application uses Infinispan as an embedded cache solution without any persitent store configurations. As I already mentioned we are constantly running out of memory. We already created some heap dumps where we could find the following in the dominator tree.

      As you can see from the screenshot there are many ThreadLocal Objects that hold refrences to AbstractJBossMarschaller objects which seems to be growing until the server runs out of memory.

      The high memory usage is mostly on the first server where we update a large cache once a day and synchronize it with the 2nd server.

       

      We couldn't find any solution for this problem yet, so maybe you can help us figure it out and fix it?

      Please find attached our Infinispan configuration.