2 Replies Latest reply on Mar 25, 2018 3:12 AM by hsepers2

    Metaspace out of memory after number of redeployments of war

    hsepers2

      Hi,

       

      After undeploying and (re)deploying a war module in JBoss EAP 7.1 a number of times we experience an outofmemory error of the metaspace.

       

      See following classloader analysis made with mat after the first undeployment:

       

      mat heapdump

       

      The classloader of our module (admin-ui) is not released.

       

      Our war file makes use of postgres sql, activemq, spring boot, hibernate.

      The postgres sql driver and activemq come from jboss modules.

       

      When I perform the same excercise with a simpler war module (a simple wicket user quide application) I experience oom errors too.

       

      Questions:

      - any ideas why for instance the poolfiller thread seams to hang on to our module classloader. And what about the activemq threads?

      - As this oom is so easily reproduced with even the simplest of applications; we must be missing the obvious. What are the guidelines concerning module usage on JBoss EAP 7.1?

       

      Thanks,

      Huub