We have a simple two-node cluster on a single server. When we restart a node to clear our internal caches, we often get the following exception when the node attempts to reload the SessionCache:
(ContainerBackgroundProcessor[StandardEngine[jboss.web]]) processExpirationPassivation(): failed passivating unloaded session
It is invariably followed by this one:
org.jboss.cache.lock.TimeoutException: write lock for /JSESSION/ROOT_localhost/PdMaBYgth15GZETt2Rl7HA__ could not be acquired after 0 ms. Locks: Read lock owners
From browsing the message in various forums, this seems to indicate that the session cache is trying to deal with some kind of rollback situation in the TransactionManager for the session cache during the passivation and failing due to a timeout. The 0ms time seems to me to be some kind of configuration issue with the session cache (we've just used whatever the default settings are).
The node comes up with no consequences as far as I can tell, but the operators restarting the node get really upset when they see a bunch of exceptions in the log, and they keep restarting the node to get rid of them. Training doesn't seem to help .
Can anyone suggest either configuration changes that might help with this situation or suggest a possible problem being caused by our web application session usage?
Message was edited by: Bob Muller
This problem appears to have gone away when we moved to using RHEL 6.2 from Solaris, same version of JBoss and JGroups. So I'm theorizing it was some interaction with the OS, but can't prove it.