We have a web application that runs on WAS 8.5 that processes requests and generate reports that are sent to print devices four times daily (ad hoc reports are also serviced by these JVMs). Reports generation are sometimes successful while at other times they failed. Failure are not associated with peak time usage because sometimes the report distribution fails at midnight when request to any of the JVM cluster member  are at minimal. Surprisely, I have noticed that when we restart all JVMs that service this application request  20 minutes before the reports runs, the reports are generated successfully without any error at least 90% of the time, althought I have seen situation where the reports still fails even after restarting JVMs but my line of thought is that whatever normally cause the issue on the JVMs might have occurred between that window of time when the JVMs were restarted and when the report was scheduled to run. Please I need help if you had experience anything with hanging threads on IBM JVM that could cause such an issue. Reviewing the log files has not been helpful. Thanks in advance.

Hi Jolomi,

  Can you share the stack of the errors from SystemOut?

 Thanks in advance.

Regards

I'd review the scope of all variables in the request. Could be related to a bad scope...

Might also  suggest a kill -3 to get a heap and look at the threads 

Have you tried running the Tivoli Performance Monitoring on it? Its baked into the console (not an add on). Take a look at the HTTP threads, as well as the JVM heap's GC cycles to see what's going on. 

Kenny Smith

Principal Consultant

IBM Champion 

www.strongback.us