On Tue, Dec 27, 2005 at 09:45:23PM -0600, Jerry Amundson wrote:
> On Tue December 27 2005 19:38, Jerry Amundson wrote:
> > On Tue December 27 2005 16:52, Jerry Amundson wrote:
> > > To move the SCM from one server to another, all volumes remaining
> > > equal, /vice/ in sync, etc. is it mainly a matter of changing the
> > > auth2 and update init script startup?
> > > I'll dig some more, too, and get the details, but am wondering
> > > about subtle things to watch out for...
> 
> I'm hesitant to even progress on this now, it seems too trivial...
> 
> http://coda.wikidev.net/What_if_the_SCM_dies
> 
> What's missing?

That's pretty much it. Coda servers are fairly dumb and the normal
client<>server protocol doesn't rely on having an SCM and such.

The only reason we have an SCM is because we don't use fancy replication
techniques to keep the data in /vice/db in sync between the servers. We
only write on one server, the SCM, and then copy from there to all the
other servers. When there is no SCM pretty much everything works except
for adding or removing volumes (VRDB/VLDB), managing users and groups
(prot_users.cdb), or changing passwords (auth2.pw).

Jan