i was thinking about adding a hook to memorypool mx bean as far as i remember it does have 'peak' memory usage for permgen and it could be charted over time. parsing gc logs is kind of hard because they will differ depending on the vm and even the gc used. 

also, the gc logs are dumped to process descriptors and not to system.out/err streams and the runner will complain about unrecognized process output.

this isn't crucial, but i think would be nice to add at some point.

On Wednesday, December 26, 2012, Michael McCandless wrote:
On Wed, Dec 26, 2012 at 4:13 AM, Dawid Weiss
<dawid.weiss@cs.put.poznan.pl> wrote:
> I think it would be nice if Mike could add permgen pool stats (mx
> bean) to his charts :) This way we would see the average permgen usage
> over time -- it's easy to spot the regression then. Something to think
> of for the future.

Hmm this is interesting!

Does this just amount to running top-level "ant test
-Dargs="-verbose:gc -XX:+PrintGCDetails" -Dtests.jvms=1" and then
parsing the GC stdout for the permgen usage?

Mike McCandless


To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org