One place where I've seen large amounts of heap get consumed is when using the Content subsystem to install a very large cumulative patch or EAR or WAR. I think we recently converted most (possibly all) places that access package bits to using streams rather than copying the all of the bits into memory. However, there may still be a few places that copy into memory. Are you using RHQ at all for installing/deploying large packages?
On 01/07/2011 02:57 PM, Bala Nair wrote:
I am seeing the following error in the server log when an agent trys to connect to the server. The server had been running for quite some time (1-2 weeks) and jconsole showed it out of heap memory. I have restarted the server with the HeapDumpOnOutOfMemoryError java option to capture anything if it happens again. Anybody seen this before?
2011-01-07 13:50:37,001 INFO [org.rhq.enterprise.server.core.CoreServerServiceImpl] Agent [aphilLabMMC1][3.0.0(59e9341)] would like to connect to this server 2011-01-07 13:50:41,350 INFO [org.rhq.enterprise.server.core.CoreServerServiceImpl] Agent [aphilLabMMC1] has connected to this server at Fri Jan 07 13:50:41 EST 2011 2011-01-07 13:50:41,351 ERROR [org.apache.catalina.core.ContainerBase.[jboss.web].[localhost].[/jboss-remoting-servlet-invoker].[ServerInvokerServlet]] Servlet.service() for servlet ServerInvokerServlet threw exception java.lang.OutOfMemoryError: Java heap space
This repeats over and over every few minutes.
Bala Nair SeaChange International
rhq-users mailing list rhq-users@lists.fedorahosted.org https://fedorahosted.org/mailman/listinfo/rhq-users