[Geoserver-users] Memory Leak after deploying geoserver.war

Hello Andrea,

thank you for the fast response and clarifying.
My Java options, concerning memory are the following.
-Xmx1568M -XX:MaxPermSize=256M -Xms1024m

I have one Tomcat instance for GeoServer and it will be the only one on that machine. The machine has 4 CPUs and currently 8GB of RAM.

One more question concerning the redeploy of GeoServer. To deploy the *.war file I start my Tomcat instance and then I copy the war file to the webapp folder. Is that the common procedure?

Thank you in advance,
Christoph

On Fri, Apr 10, 2015 at 1:28 PM, Christoph Kleih christoph.kleih@anonymised.com wrote:

Dear Geoserver Users,

I have deployed a new geoserver.war in a new tomcat instance. Currently I cannot work with it, because my http/ajp mapping isn’t configured correctly.
Besides that, I get the following output in the catalina.out log. (see the log below)

There seems to be a problem with a memory leak and the Threadpool.

Can you give me a hint what is going wrong and if that is a problem.

It’s a known issue without an easy solution, GeoServer is a complex, large web applications
that uses a set of technologies which weren’t exactly meant to be used in a webapp context (emf, jai).
I’ve tried to fix those errors spending a few days some years ago, without any luck, and in time,
more of them piled up (since nobody managed to solve them all, and it does not make a difference
if you get 1 or 100 of them, you still end up with a permgen leak).

The usual approach is to deploy GeoServer in its own private Tomcat, and shut it down when
you need to redeploy. Not as bad as it sounds, considering that GeoServer will force you to
increase the permgen settings anyways (at least 128m) and under load it ends up eating a
good portion of your heap (each WMS request allocates a few megabytes of heap for the
drawing surface, large ones can reach the tens of MB, by the time you have some concurrent load
using a couple GB of memory is not uncommon)

Cheers
Andrea