On Apr 25, 2007, at 10:31 AM, Dario Andrade wrote:
That has to do with the PermSize. It is the classloader's memory.
Set -XX:MaxPermSize=256M at startup.
--
Dario
Jochen Zink escreveu:
Hello,
I'm using Geronimo 1.1.1 and have a problem.
I'm developing a little JSF-Webapplication. First, I deploy the
WAR to geronimo and after modifiying the sources I redploy the
application WAR. That works fine. But after a while, I get
OutOfMemory Exceptions while redeploying. After restarting
Geronimo, everythink is fine.
Is this a bug, or a kind of "wrong use of redeploy"?
Hi Jochen,
You are using redeploy properly.
There are different types of OutOfMemoryErrors. Your scenario
suggests that you are running out of PermGen space (e.g.
"OutOfMemoryError: PermGen space"). PermGen space is used by the Sun
JRE to store loaded Classes. As Dario suggests, configuring the
MaxPermSize to a larger value (default is 64M), may avoid the
problem. However, if there is a ClassLoader memory leak, with each
undeploy/deploy (e.g. redeploy), the amount of PermGen space being
used will continue to grow. Increasing the PermGen size may only be
postponing the inevitable OOME...
ClassLoader memory leaks may be caused by application code or a
Geronimo bug. We did test for Classloader memory leaks in 1.1.1 and
I'm not aware of any bugs.
I suggest you increase your MaxPermSize to 128M. If you continue to
see a problem, then you'll need to dig a bit more. A good Java
profiler is pretty useful in tracking down these problems. If you can
make a testcase available, then probably someone in the community can
take a look... (if you can recreate the problem on one of our 2.0
milestone drivers, then we'd definitely take a look... :)
--kevan
