The tricky parts about multi-homed file servers are volume operations
and callback state. The tricky part about multi-homed clients is
callback state. To truly solve the callbacks problems we would have
to either rev the protocol or make some nasty hacks using DNS or
something even worse which I don't want to think about right now.
Just something for you to think about as you try to set up a
multi-homed configuration.
I can't tell enough about your configuration from the post to
info-afs, but I think you could do something roughly along these
lines:
Given a server with IP addresses EtherA and FiberB, ALWAYS create all
volumes "on server EtherA". Then make sure that your clients have a
route installed in their routing tables to EtherA via FiberB.
If you have no other router on this subnet but use the server as a
router, and you run RIP (for example), this will happen automatically.
If you do have another router on this subnet, you may have to
explicitly install the route. Also, be aware that if either the
router OR the fiber interface on the server goes down, the client will
almost certainly use the other route to the server (via the Ethernet
interface) and the callback state on the clients on the same subnet
will become inconsistent. OTOH, if the Ethernet interface on the
server goes down, all the other clients in the world won't use the
server's FDDI interface...
Lyle Transarc 707 Grant Street
412 338 4474 The Gulf Tower Pittsburgh 15219
