On Thu, Jan 06, 2011 at 06:00:30PM +0000, Hofmeister, James (WTEC Linux) wrote: > > In the RH436 class it was verbally discussed during the fencing > chapter that it was a bad thing to shutdown a node, but to instead > power down the machine which has been implemented in many of the > fence_ scripts. > FWIW, I suspect the context of that comment is different from that of your question. In the context of fencing, the cluster has already decided that the box being fenced is in a bad state, and should go down as quickly as possible. Seeing that the box is being fenced, it likely means that it's not communicating with the cluster, so when an orderly shutdown tries and stop the cluster and associated services on the node that's being fenced, it will hang, and the shutdown won't continue, so the fencing won't be successful, and so your cluster services like GFS will remain hung. Better to power kill the node so that the cluster can continue as soon as possible. -- Hugo Lombard -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
