Re: HELP - What is this OC4J thread dump telling me!

Ed <ed.barrett_at_bigfoot.com> wrote:

> I am really hoping someone has the knowledge and a spare 10 minutes to
> tell me what this thread dump is telling me - an explanation of how
> whatever is determined would be appreciated but we are in "result"
> mode at present.
>
> We have an application running on OC4J standalone 9.0.3, JDK1.4.1,
> Solaris 8.
>
> The application works well under low load but once we simulate tens of
> users the JVM deadlocks - what I mean by this is nothing responds, the
> CPU goes down to and stays at 0% and the only way to get things back
> is to kill and restart.
>
> Normally kill -3 does not work in this situation but I have persevered
> and eventually got this thread dump. Now all I need to know is how to
> determine where the (dead)lock is ... trouble is the stack trace
> reading articles on the web don't seem to apply ;-(
>
> BTW, have tried 1.3.1_06 and _07 and all exhibit the same error.

The dump looks particularly nasty... Especially the com.entegrity.* classes, I presume they're part of some authorization/access control framework, aren't they?

Have you been able to determine the cutoff point for the deadlock? In other words, if N is the number of simulated users with which the JVM locks up, how high is N? My theory is that you may be running into a couple of problems:

• bugs or misconfiguration in the com.entegrity.* classes
• OS limits

Can you deploy your application on another OS, but using the same JVM? Perhaps testing with a 1.4.1 JVM on Linux or even Windows may help in determining the root cause of the problem.

Good luck, if you end up finding a solution I'd appreciate a summary if possible.

Sebastiano Pilla Received on Sat May 03 2003 - 15:29:08 CDT