Problem with jboss on Linux

[sreram]

We are using JBoss 3.2.1 integrated version with Tomcat and we are facing peculiar problem that application crawls after 5 hours and stops serving request. The JVM, however still remains alive. We are using Linux cluster with session replication.

We see that memory is being used to 99.8% and not being released. We ran JProfiler also, but from application angle, it is not showing anything abnormal. We are just wondering if this is a commin trend and pattern in the open source application servers - or anything typical to JBoss.
For both the questions, the required skill set would be - Open Source Application Server preferably JBoss, J2EE, EJB, Linux, Tomcat Servlet engine.

[rajaboobalan]

Do you have any feedback on this issue? We are planning to go production with LINUX /JBOSS 3.2.2 in a clustering env. I wish to get a feedback ASAP.
thanks

[mikefinn]

FWIW - this is configuration (Linux/JB 3.2.2) what we are on in production. We have Apache/JK2 in front of two 3.2.2 cluster nodes, using sticky sessions, and it's working well. The only problem we have is that automatic/seamless failover of sessions doesn't seem to work, as it seems the subject/principal on the session does not get replicated with the session. It just means that, if a node goes off the air, users will see a login screen on the next page request, for the node that JK2 sends them to.

Re: the memory problem in the first post above - what do you mean by 99% utilized? System RAM? Stack? Seems like a leak somewhere. It's surprising a profiler can't find it.

Mike