We have roughly 15 jboss apps all part of the same cluster. We've noticed that when we are doing a deploy and stopping/restarting a subset of the cluster (say, 5 servers) that our database will basically get locked out (lots of jboss transaction timeouts, a few database transaction timeouts). Eventually, it always sorts itself out but can take from 1 minute to 5 minutes. If we stop just one node and restart one node at a time, we never see this problem in the cluster.
I can't find any issues related to this behaviour in this forum...is this a known bug, or the first time this has been brought up?