On Fri, Nov 23, 2001 at 07:51:26PM +0000, dan.lynch_at_fensystems.co.uk wrote:
> However, I seem to be having some problems setting up replicated servers
> reliably. The primary installation goes smoothly, but upon power cycle of
> all of the servers (though I seem to have narrowed this down to the SCM) I
> cannot access the mounted filesystem (/coda). I can clog ok, cfs lv /coda
> says that the volume is up and connected, but I get "connection timed out"
> upon an ls /coda. Both replicated servers are running the necessary
> services and I get the same message even after a system purge of vice
> (what can I say - a fit of despearation!)

Ehh, I guess one of the servers is complaining. The 'cfs lv' probably
only goes to one server who is happy, but the other server has some
problem so the 'ls' which talks to both servers gets an error and drops
the connection.

Check the server and client logs, and try to get information about the
rootvolume and it's underlying replicas using "getvolinfo <scm>
<rootvolname>", "getvolinfo <scm> <replicaid>", etc.

> Also, in a semi-related question, is it possible to add servers to a
> replication group AFTER the creation of a replicated volume?

Not yet, the client 'almost' supports it, we just need to hook into some
callback to force a revalidation of the cached volume information (i.e.
trigger a ViceGetVolumeInfo call and update the local state). I haven't
looked at the servers yet, but they will require more surgery to deal
with resizing a volume storage group.

Jan