VCS will normally prevent SAN disk mounting on both sides, this will only fail if you loose both heartbeats at the SAME time (which shouldn't happen as they should be independent) or by administer error (like if you online VCS on one node while filesystem is mounted on node not running VCS AND you override Autodisabled flag).

You can protect against split brain using CPS, but this would require 3 extras servers, but you can add CPS to existing clusters (windows as well as UNIX I believe)

If you use Solaris SVM or ZFS, as oppose to Solairs slices, I would guess (don't know much about Solaris O/S products) that these would give you some protection against mounting concurrently.

Mike

Also confirm that filesystem in Mount resource does not exist in /etc/vfstab in any of the cluster nodes.

This topic in VCS Admin Guide applies to SF as well as non-SF data:

http://www.symantec.com/storage-foundation-basic/data-sheets-white-papers says:

Veritas Storage Foundation Basic is a free-license version of Veritas Storage Foundation, specifically designed for workloads running on single and dual-processor servers — including edge-tier, departmental, and test/development systems

and

Veritas Storage Foundation from Symantec product designed for heterogeneous online storage management of edge-tier workloads with up to four file systems, four volumes, and two processor sockets per system.

and you can download whitepaper from here too.

I believe everything else is the same, so you can create huge filesystem, extend online and you can use DMP.

Mike

Also I would make Mount resource depend on IP resources as this protects against the administrive error mentioned in my first post as if you try to online the service group and the IP is up on the other system, the the IP resource will fail and the mount will not happen.  This also protects you against losing both heartbeats at the same time, so long as the NIC with IP in servicegroup stays up.

Mike

That sounds like an odd application  - if you stop the IP first, before the application, then all connections are suddenly ripped away and so the applicaton cannot close the connections cleanly and if you have to stop the IP first, then how does the application operate when it is not in a cluster, because here you would normally use the IP of the server which is there all the time and never goes down.

Mike