Originally posted by: SystemAdmin

Was curious if anyone has had any experience with gpfs 3.2, and you are using the tie breaker disks for a 2 node cluster. Need to know if I setup the cluster with both nodes as quorum nodes and can I just setup 1 nsd disk as a tiebreaker disk or should I setup more than one.

We are running aix 5.3, gpfs 3.2. 2 nodes with SAN attached storage to both nodes. It will basically be a file server for a application.

Originally posted by: MisterX

We havent upgraded to 3.2 yet, but here's my 3.1 config:
Configuration data for cluster cap_plan.psccapsip01:
<hr />
clusterName cap_plan.psccapsip01
clusterId 11173460535293426351
clusterType lc
autoload yes
useDiskLease yes
maxFeatureLevelAllowed 911
minMissedPingTimeout 60
tiebreakerDisks gpfs1nsd;gpfs2nsd;gpfs14nsd
psccapsip01
takeOverSdrServ yes

File systems in cluster cap_plan.psccapsip01:
<hr />
/dev/gpfslv

Originally posted by: SystemAdmin

Thanks for your response. Is the cluster a 2 node cluster and are both nodes setup as quorum nodes? or is it just 1 node that is a quorum node

Originally posted by: gcorneau

You want to set up both nodes as quorum nodes. Otherwise, if the single quorum node goes down, you lose the GPFS file system access on the remaining node.
FYI, the tiebreaker disks do not have to be dedicated to the TB function, they can be regular file system disks.

I would also suggest using 3 tiebreaker disks. If you use just one TB disk, and that disk has any problems, then you lose the TB function and the file systems unmount. You want an odd number of TB disks and GPFS supports up to 3.

Hope this helps.

<hr />
Glen Corneau
IBM System p Advanced Technical Support