Java.lang.OutOfMemoryError solved (was NOT a leak)

Question

Alright I'm just posting this to hopefully save someone some time in the future.
This was a bug that dragged out for over a year. An application would be deployed to a production app server and low and behold OutOfMemoryError's would start to be thrown, naturally the new application would look like the most likely culprite. So now the unlucky developer gets to spend hours running the memory profiler and looking in vain for a memory leak that does not exist.
The actual problem was not the Heap running out of memory as in most memory leaks. No, the problem was the app server runs out of Perm Generation space (this is where class loaders and reflection data is stored) by default the Max Perm space is 64MB which is fine for most apps, but on an application server (OC4J in our case) when you have 10 apps with tons of JSP's this space can fill up.
You can check out your perm gen usage by running the jmap tool with the -heap option and pass the pid of the oc4j instance to it, this will tell you your max perm gen space and your current usage.
Anyway to up your max perm gen space, start the java process for your app server with the parameter -XX:MaxPermSize=128m
Problem solved.

Answer 1

Good Luck,
If 10.1.3 uses 1.5.x of the JVM then I believe they added an extra string on the end of of the outofmemoryerror that says Heap or PermGen so you have that hint to go on.
You should take a look at the jmap tool in 1.5 though, you can make it print a dump of what's in the heap then open it up in the tool jhat to take a graphical look of what's hanging out in memory, very useful.
By the way, thanks very much for sharing your
experience.
I will surely encounter your problem since I have a
big app.
I begin deployment to OAS 10.1.3 at the end of the
month, I waited for the latest version since it is
much easier to maintain then previous versions.
Regards
Fred