On Fri, Jun 14, 2002 at 10:30:55AM +0200, Stephan Stapel wrote:
> Since replication seems to be done automatically, coda seems to be quite 
> suitable for this task. However, I read that I have to set up exactly _one_ 
> SCM machine on my coda cell to control it. Doesn't this set up introduce a 
> single point of failure? I mean, what happens when the SCM machine is absent 
> and/ or needs to be replaced by another machine, perhaps a new one?

All Coda servers are identical. The only 'special' property of the SCM
is that all other servers poll this machine for updated user, password
and volume databases.

As long as you don't add/remove users, change passwords or create/destroy
volumes there is absolutely no need to have an SCM.

When the SCM is totally dead, any other Coda server can take over the
role. You just need to inform all the other servers who the new SCM is.
This is done by putting the hostname of the new SCM in /vice/db/scm and
restarting the updateclnt daemon (/etc/init.d/update.init restart) on
all servers.

Jan