On Tue, 2009-04-28 at 14:53 +0200, marco perugini wrote:
> Dear all I have a little question, about a real scenario.
> I have this in my network:
>   - A database cluster, call A, where live master instance of Slony;
>   - A database cluster, call B, where live slave instance of Slony;
>   - A simple database, call C, where i'd like to have a replication.
>   
> So, at now A is master and replicate a schema (call S1) over the database B, the slave. 
> I need to create another schema (call S2) in B that must replicate over the simple database C(S2 schema slave).
> My question is:  can I have a Master process of Slony (for instance in B) where is live a slave process?
> If yes, how can i configure it. If No, how can I do with this environment?
> Thanks everybody for help.
> Regards,
> marco

Yes, you can do this.  You would use one set for S1, making the node on
cluster A the provider, and Cluster C a replica.  You would create
another set for S2, making the node on Cluster B the provider and
cluster C the replica.  The same set of slons will service this cluster.

A couple words of warning with this configuration though 

-I'm not sure if you want to have Cluster B also be a replica of of the
set  S1.  If this the case, you will not be able to rely on pg_dump for
dumping cluster B if you are using a version of Slony prior to 2.0.x.
This will be a mixed cluster (a provider and a replica), the catalogs
are corrupted (by design) on a replica.

-We have found the multiple provider clusters to be highly error prone.
There is no problem when replicating data (we've never had a problem
with regular operation), but config changes often don't behave as
intended, and often result in lost nodes.

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