We're on JB3.2.3 with Jetty 4.2.17, after a recent upgrade from 3.0.8. We have a cluster of 3 machines, which works well in terms of failover - sessions are replicated as expected.
However, these machines (dual processor Xeon, 1G RAM) see increased memory usage to the point of needing to restart JBoss several times per week to release the memory used - then the cycle starts over.
We only see this increased memory usage on the clustered machines, and not on non-clustered machines.
My feeling is that the http sessions are not actually being totally released when the session is removed. I see the removal of the session in the loag, but it seems the memory remains.
We have changed some of the session timeout values in our web.xml, but I'm not sure if there are other values that need changing someplace else that the cluster is using - it certainly is possible that everything is functioning correctly, and we need to change a config...
Have others noticed this same issue?
What parameters should we be looking at that affect clustered session times?