Looking over the documentation for updating the XCS code on an EVA6100,
it seems they have a procedure for doing it live. This is a very nice
thing, as taking everything connected to this array down for the duration
of the patching is a major event.

It seems the way they do this is to set the driver/FC-BIOS to have a
port-timeout of over 30 seconds. So it doesn't declare a path broken
until the 30 seconds has passed, presumably buffering I/O until the path
becomes available again. This gives time for the EVA controller(s) to

It's a nice idea in principle, but I see a big one for Novell Cluster
Services. The storage heartbeat / split-brain-detector process will fail
during such an outage, even with the port timeout set high.

Would putting the cluster into Maintenance Mode allow it to survive this,
or do we have to expand a heartbeat tolerance somewhere?

Novell, it does a network good