On Tue, Mar 20, 2007 at 10:30:15AM +0100, Enrico Weigelt wrote:
> I just had another problem w/ the coda client: it tells me the 
> server was down. cfs checkservers hung for a while and then said 
> the server still was down. But the server was up and ready.
> 
> After restarting venus, evrything worked fine.
> 
> How can I get around such errors in future ?

Not sure what the problem is/was so I don't know how you could avoid it.

cfs checkservers hangs while the client is sending probe requests to the
server. If the server doesn't respond to a probe we try an additional 4
times over a 60 second period.

Server 'liveness' is based on having an established callback connection.
So as a result of the probe request, the server tries to set up a
connection back to the client and send a null-rpc operation over the
callback connection.

So either your server was not receiving the probe requests, or it failed
to set up a reverse connection to your client. This can sometimes be
caused by masquerading firewalls, although we always try to use the same
port pair between the client and the server for both the forward and the
reverse connections which works for most firewalls. The only cases where
I found that it doesn't work is when the firewall only accepts the first
reply packet and drops any remaining ones, or when the server is
multi-homed and is sending it's response from a different ip-address.

Jan