HAWKER, Dan wrote: >>> I am not running any cluster software per-se, I am only >> using CS to be >>> able to gain the shared clusterFS (GFS). >> GFS is cluster software. "per-se" or not ;-) > > Point taken :) was just hoping that this procedure was just for application > clusters. > >>> Is this still the case (need to halt cluster services) or can I >>> accomplish this somehow??? >> It is still the case that you need to remove the two_node >> flag and restart the whole cluster. > > Grrr. Ah well, if that's how it is... > > Presume by restarting the whole cluster, you mean unmount all filesystems > and restart the cluster. > > Could I just remove one (quiet) node from the cluster (presently not doing a > great deal) and then update the file/restart the cluster services on the > single busy/live one??? Its more the umount of the fs that's the problem (is > in continual use)... > > Guess I'll be planning a quick up and down of the cluster in the next day or > so. No, you can't do any sort of a "rolling upgrade". Once you change the config file on one node, the node that's left in the cluster will refuse to let it join because the config file versions differ. It's an unfortunate consequence of the fact the two-node clusters are a special case. Normally a cman cluster requires (n/2)+1 nodes to keep quorum. patrick -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
