On Tue, 2006-12-05 at 07:50 +0100, Pena, Francisco Javier wrote: > > On Tue, 2006-11-28 at 16:20 +0100, Pena, Francisco Javier wrote: > > > Hello, > > > > > > I am doing some tests with a GFS cluster in our lab. As part of the > > > tests, I am trying to upgrade a two-node GFS cluster to > > three nodes. > > > This usually requires shutting down the cluster, but if we > > use qdisk I > > > supposed this was not true anymore... > > > > Unfortunately, that's not true -- because cman flipping from > > 2->non-2 node mode requires a restart of cman, which qdisk > > can't control nor fix. :( > > > > -- Lon > > > > Ok, my fault then. I assumed that removing the "two_node=1 > expected_votes=1" options from the cman configuration (as suggested by > Josef Whiter in a previous thread) was enough to prevent cman from being > restarted in this scenario. It should be, but CMAN doesn't know how to change between the modes, unfortunately. > So, if I had a three node cluster and wanted to upgrade it to four > nodes, would the qdiskd number of votes be properly propagated to all > nodes? To go from 3..4..N with nothing special. * Add the new node to the configuration on the running cluster (all 3 should be running). * ccs_tool update /etc/cluster/cluster.conf * cman_tool version -r (new_version) * To be paranoid, I'd use 'scp' to copy cluster.conf to the new node * Start new node The *only* time CMAN gets weird is going 2 to anything more than 2. If you were using qdisk in a 2-node cluster and *did not* have the two_node="1" set, you could probably go to 3+ nodes without a problem, but I haven't tried it. -- Lon
Attachment:
signature.asc
Description: This is a digitally signed message part
-- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
