I am running Jboss 3.2.0 (clustered using 3 HAPartitions). What I am seeing is, failover works without a problem, but when we bring the failed server back up into the cluster it seems to join the cluster before its ready, and when the client connects, the client ends up getting an error, even though there are other servers ready to service the client. If I wait for the server to come all the way up before sending the request through the client, This problem doesn't seem to happen.
Is there a configuration paramater I have overlooked to ensure the node doesn't join the clusters until it is 100% up and running? I have the for-pay docs and haven't run across it yet if its in there.
Can you please try with 3.2.2 (out tomorrow) and if you get any exception, post them here.