I'm experiencing the same problem that was reported in issue #5106.
I'm using Glassfish v2.1 on CentOS 5.3 x86 on VMware 3.5 ESXi with a
HW-balancer. Simple roundrobin without persistence. (I need without persistence
because most of the clients work behind a NATed router)
Everything works (and i can see the messages if I configure the logger with
FINE) with clusterjsp app.
When i shutdown one instance and that start it up again but the instance
restarted always create a new session.
I can see that the restarted instance resquests (broadcasts) the session id but
it doesn't receive a response and creates a new session.
I can confirm that it broadcasts the request using tcpdump.