Hi All, I am facing a hang while shutting down my cluster systems. The type of shutdown used is a "cluster_shutdown". The servers in cluster are Itanium servers. I am aware about the cluster nodes waiting for each other to come to common execution point and shutting down together. But this is not the case here. One of the nodes hang much before coming to this common execution point, which is creating complete cluster hang. I am trying to figure out why this particular node always hangs while shutting down. Can anyone point out the possible reasons of one node in a cluster hanging during shutdown?

What is different between the nodes in the cluster? Is the "offending" node running anything that is NOT running on other members of the cluster?

Is a sys$manager:SYSHUTDWN.COM present on the node ? If yes, put a "$SET VERIFY" at the beginning, and watch if shutdown 'hangs" executing this file, or answer NO to the question "execute site specific shutdown procedure?" id running shutdown interactively, or the respective option in SYSMAN SHUTDOWN NODE /NOINVOKE_SYSHUTDOWN to see if something in the site specific shutdown makes it hanging. Joseph Huber, http://www.huber-joseph.de ITRC

I'm not quite sure if it is necessary, but also add the REMOVE_NODE option to the shutdown command. It could be that a non-voting node hangs during shutdown because cluster quorum has no longer sufficient votes. Joseph Huber, http://www.huber-joseph.de ITRC

I would actually discourage the use of the "REMOVE_NODE" option. That's really only intended for instances where you're shutting down just one node, and that node is expected to be out of the cluster for an extended period of time. I understand why you're suggesting this, but that's the whole purpose of the "CLUSTER_SHUTDOWN" option - to prevent quorum adjustments until all nodes are ready to (gracefully) exit together. I suspect Bill's on the right track, although I've seen occasional instances of network shutdown hangs, and this could be at play here. Joseph's SET VERIFY suggestion may help pinpoint where the problem lies. Use of the CTRL-T feature may also give you a clue if you're running the shutdown interactively on the problem node. Are you using a console management solution? Having a log of the messages displayed on the console terminal during the shutdown may also help narrow things down a little.