On Wed, Aug 22, 2007 at 05:00:19PM +0200, Brieseneck, Arne, VF-Group wrote: > Hi all, > > Has anybody an idea if it would be possible to fence first via a > controlled shutdown second fencing via fence_impi? > The advantage would be, that the filesystem of the fenced node have a > better chance to be intact after shutdown then power cut off. The disadvantages are: (a) it increases cluster recovery time by *minutes* at the very least (how long do you wait before cutting the power?) (b) it might hang during shutdown, seemingly responding to the shut-down request but never actually reaching 'off' state... (c) the node - if it's actually having a problem - could still write to the disk while we're trying to fence it. Some could argue that this data might be bad (particularly if the node's problem is a bad memory stick), and lessening the window for this is a good thing. There are also some IPMI implementation-specific problems (bugs?) preventing this from being possible. -- Lon -- Lon Hohberger - Software Engineer - Red Hat, Inc. -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
