09-04-2008 05:46 AM
We are using SAN architecture
since several years and a lot of clusters (Solaris, Linux, Oracle RAC, etc..)
are connected without any trouble.
met a problem with 2 Clusters under Linux RH4.
reboot one node, the other loses it paths to disks. The two nodes and the
storage are in the same Zone. If i create two separated zones instead of one,
this trouble does not occur anymore.
have met the same problem ? and/or is able to explain this phenomena (LIP reset?)
09-10-2008 10:50 AM
Although I have limited understanding, since no one has replied I give it my best shot.
By zoning the servers you are preventing a LIP reset caused by server reboot to reach the other server in the cluster. As a best practice you should always have single initiator single target in one zone. Maybe the RSCN's are causing the multipathing driver to bring down a path, and when u change the zoning scheme, the RSCN's no longer affect the hosts.
RSCN = request for state change notification.
Although this shd not be the case, some multipathing software behave crazy.