On Fri, 2008-01-11 at 15:30 +0100, Michael Weber wrote:
> > Lets say that the external database you are replicating to your
> > Institute is A.  The internal database that you are replicating into in
> > your Institute is B, and the the third database in your Institute is C.
> >
> > Set it up like this:
> >
> > Set 1 - A is the origin
> >         A is the provider to B
> >         B is the provider to C
> >
> > Create Set 2 with the tables and sequences local to your institute
> >
> > Set 2 - B is the origin
> >         B is the provider to C
> >
> > Now, there is one big caveat to this setup.  You will no longer be able
> > to use pg_dump to backup database B.  You can not pg_dump a replica, and
> > what you've done with this config is make B both a provider and a
> > replica.
> >
> Thanks for the quick reply!
> 
> Once everything is running I don't think a dump will be necessary.
> What do I do if things break (e.g. server C dies) and I need to refill
> the database? How do I do this without a dump? Or is pg_dump no
> problem with slony stopped.

You would use slony for this.  If you lost C, you would drop node C from
replication, then re-add and re-subscribe it.  This will copy all the
data from the sets over to it.

> What do I need to take care of when I change the configuration? Just
> stop slony, change the config, probably run initialize node, and then
> start slony again? Is it that easy?

You don't need to stop slony, you can make the changes on the fly.  I
would recommend (as always) testing the changes out before doing them,
especially if you are new to slony.

-- 
Brad Nicholson  416-673-4106
Database Administrator, Afilias Canada Corp.