On Thu, Mar 4, 2010 at 4:23 AM, Jeff
Karpinski <jeff 3d0g net>
assigned Red Hat engineer was on-site today and pointed out the
blindingly obvious solution. Can't believe I didn't think of it: Run
NFS as a clustered service and have the VMs mount that. That way ANY
system - even outside of the cluster - can also access the data.
This is what we are doing, works great. We considered presenting the
raw devices from our SAN (fc connectivity instead of iSCSI) to the
VM's, but opted against it due to dynamics of changing # of VM's and
GFS requirements for journals / # of nodes, as well as multicast issues
(each dom0 uses a different routed network for VM's). Each VM mounts
NFS from its host.