I am more than happy to share my findings once I get through this mess.
The
problem is debugging Java apps is not my strongest point, so just trying
to
make sure everything is setup properly. Also I have no problems on dev
or
qa environments which is even more strange, seems to only happen in
production. Could be because heap is allocated larger and runtimes is
larger, so maybe something is getting blown there. Cant wait for Rails
2.2
Will certainly post my findings and any good java heap debugging tips
are
always welcome so I can hopefully help others going through this…
Hi AD,
You can add:
-XX:+PrintGCDetails
then You will see what acctually GC is doing.
Also add
-XX:NewSize=150m -XX:MaxNewSize=150m
Huge NewSize can kill your app, had this once on java app.
It doesn’t gcing for a long perdiod of time (huge new)
and then GC can’t find continous space to put objects to the next space,
because next space is already clustered.
Another thing i noticed is that Your jvm uses ParalelGC,
You can experiment with other GC algorithms.
Thanks Charles. I am getting jprofiler right now, anything specific in
JProfiler I should look for? Debugging Java Heaps is unfortunately not
my strongest point.
I believe JProfiler can track objects back to their roots; look for
things like org.jruby.Ruby or ThreadContext instances that are sticking
around; those are typical signs of leaking threads or runtimes.