Coda File System

Re: Configuration for replicating servers

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Tue, 4 Jul 2000 21:30:41 -0400
On Tue, Jul 04, 2000 at 11:22:24AM -0700, Phil Nelson wrote:
> 
> >- Make two new entries in the /vice/db/VSGDB file. One for your new server
> >  by itself, one of the form: E0000104 scm-server second-server. 
> >I don't understand the 'One for your new server by itself' part here.
> >I don't know what volume address to use, as my 2nd server does not have
> >a root-volume of its own. Do I have to use the volume address of the
> >root of the SCM? Or do I have to use a fake number?
> 
> The VSGDB defines "volume storage groups".  When volumes are created
> via "createvol_rep", you specify a volume storage group, by number, for
> where the volume is to be stored.  The numbers appear to be arbitrary (now)
> but have historically started with E.......  Given your srv1 and srv2,
> your VSGDB file might look like:

There is actually some code in the clients that relies on the VSGDB
numbers to always start with E0. These numbers are hexadecimal encodings
of multicast ip-addresses, although Coda does not use multicast ip.

> To help you here, we need some server logs.  I suspect that you have
> run into a bug that has plagued me for some time ... a volume with
> only one replica (that is stored only on one server) is running out of
> room in the resolution log.  If this is the case, one possible fix is

That is very likely if the COP2 messages are missing. This happens often
during (weak-)reintegration, and in normal cases the resolution logs are
then truncated by a following resolution. However single replicated
volumes do not ever develop a server-server conflict, so resolution is
never triggered.

Jan
Received on 2000-07-04 21:47:09