Coda File System

Re: Restart replication after one Server was down

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Sun, 7 Oct 2001 03:30:19 -0400
On Sat, Oct 06, 2001 at 09:29:34AM +0200, Matthias Teege wrote:
> I have two coda servers and one of them, not the SCM, must I
> shutdown yesterday. I have an replicated volume and after the
> restart it looks like that the replication didn't take place.
> 
> There are no errors in the server logs. How do I restart the
> replication?

When the available volume storage group changes, i.e. the server comes
back up, Coda clients invalidate all cached objects. As a result
whenever any object is accessed the client will refetch the attributes
and version-vectors of all cached objects. When the VVs are different
the client triggers the server-server resolution that will bring both
servers back in sync.

So to quickly force both replicas back you simply run 'ls -lR' in those
replicated volumes that might have missed updates (once the client knows
both replicas are available). Or just ignore the temporary inconsistency
as soon as a client actually accesses any file or directory it should be
brought back in sync automatically.

Jan
Received on 2001-10-07 03:32:01