Transactions fail in a cluster node after it gets reconnected to the cluster

[bozhidar]

I have a cluster setup consisting of 3 JBoss 5.1 nodes. Sometimes a cluster node gets suspected of being dead, because of a long running garbage collection (around 50 seconds) on its JVM, by the other nodes. After the suspected node gets responsive again, the cluster view gets up again, however no transactions worked after that.

I've attached some short log files from all the three nodes. The last node is the problematic one.

I have absolutely no idea what might be causing this and I failed to find information relevant to my problem online so I'd like to ask you for your help. Thanks in advance!

[wdfink]

Do you have UserTransactions, mean that a Tx is spread over two nodes?

What EJB version do you use?

[bozhidar]

No UserTransactions in the application - everything is container managed. I'm using EJB 3.0

[wdfink]

How the Tx is started? A MDB or an explicit call to SLSB's?

Maybe you set the logging level of 'com.arjuna' and 'org.jboss.tm' to TRACE, you might do it after the node is in trouble (it will take up to a minute to pick up).

Maybe it will show what happen (you might check and compare the same log if the server is correct running).