Suppose we have a machine in a JBoss cluster that needs to be brought down because of a hardware failure (eg. a CPU or memory chip needs to be replaced asap). How can we:
Remove the bad machine from the partition so that no new clients try to use EJBs on that server instance.
Figure out when all clients that are currently connected to that machine have finished running.
Does anyone have experience with this scenario?
I don't know if you've already got an answer to this on your own but here's my reply anyway:
It's easy - just shut down the jboss on the faulting machine. Clients using that machine will fail over to the other nodes in the cluster automatically. This is something we do a lot - like maybe once a week. There's no need to remove it from the partition or anything like that.