[Linux-cluster] GNBD Configuration Question

[Curt Moore]

Hello all.

I've been experimenting with RH Cluster Suite and GFS and have come
upon a few questions which I hope the list will be able to help. 
Kudos to all of the developers, RHCS and GFS are really cool!

To my question, I'm trying to setup a storage network with GFS and
GNBD using a 3 layered approach as shown in Figure 5 in the following
link:

http://www.redhat.com/magazine/008jun05/features/gfs/#fig=multipath

and also here:

http://www.redhat.com/docs/manuals/csgfs/browse/rh-gfs-en/s1-ov-perform.html#S2-OV-MODPRICE

Obviously, the intent is to eliminate any SPOF for the storage network.

For the sake of example, let's say that I have 2 GNBD servers
connected directly to the SAN, snode001 and snode002.

If I export the the same SAN block device from these 2 GNBD servers
for load sharing purposes and snode001 fails, how do the GNBD clients
importing that block device from snode001 know that they can also find
that block device on snode002?  Is this somehow handled at a lower
level by configuring a resource within the RH Cluster Suite?

I've scoured the list archives and found the following example, using
multipath, which seems to come very close:

http://www.redhat.com/archives/linux-cluster/2005-April/msg00065.html

However, the documentation states that multipath GNBD cannot be used
with GFS 6.1:
http://www.redhat.com/docs/manuals/csgfs/browse/rh-gfs-en/ch-gnbd.html

Is there another way of accomplishing this without using multipath or
am I misunderstanding the concept of how multipath is utilized in this
setup?

Any feedback would be appreciated.

Thanks!
-Curt