Veritas Campus-Cluster -- CP Disk/Server single point of failure?

[eckerc]

Hello,

I'm planning a Veritas 6 two-node campus cluster using RedHat 6.

We have two sites, so every node operates in a separate site.

Both sites are in the same LAN and SAN.

As Tiebreaker I want to use three coordination points.

One coordination disk (eg. SAN-LUN) on each site of the campus and another coordination disk (eg. iSCSI Target) from a third side.

I read the Administrators Guide (page 624) and noticed, that the nodes at both sites commit suicide, if the CP on the third site fails.

Is the third Coordination Point on the third site a single point of failure?

What happens if the iSCSI Server (on third site) crushes?

Thanks,

Christian

[eckerc]

I also read it is possible to use only one CP Server als Coordination Point.

What happens to the Nodes, if they can communicate but the CP Server on the third location goes down?

[avsrini · Accepted solution]

Hi Christian,

 You can use Co-ordination point server, instead of using iSCSI LUN as 3rd CP.

With 6.0, you can have multiple IP addresses configured in multiple NIC's on CP server,

avoiding Single point of failure.

CP's or fencing disks are only used in-case of cluster state changes. i.e, if there is a

LLT link failure between cluster nodes or one of the cluster node panic's. Also in split-brain

situations, to race for CP's to survive. So if the nodes are communicating ok, CP's are

not used.

Hope this clarifies some doubts.

I would suggest to read VCS 6.0 guide for fencing to know more about CP's and fencing.

Regards

Srini