Thanks again Christopher, this is now real clear to me... 


J?r?mie.


> -----Message d'origine-----
> De : Christopher Browne [mailto:cbbrowne at ca.afilias.info] 
> Envoy? : jeudi 4 novembre 2004 18:48
> ? : J?r?mie LOPEZ
> Cc : slony1-general at gborg.postgresql.org
> Objet : Re: [Slony1-general] One cluster for all 
> replications... or not?
> 
> J?r?mie LOPEZ <jlo at mgn.net> writes:
> > Thanks a lot Christopher, a complete example is sure pleasant.
> >
> > However, my case is different, so I'll be indeed more specific:
> > Let's say I have 2 PostgreSQL servers - PGSRV_1 and PGSRV_2 - and 3 
> > different databases - DB_CLIENT, DB_PRODUCT and DB_ORDER.
> 
> If there are three databases, then the answer is clear: you 
> need three clusters, and hence six slon processes.
> 
> > DB_CLIENT is hosted on PGSRV_1. DB_PRODUCT and DB_ORDER are 
> hosted on 
> > PGSRV_2.  I want to set up replication of the databases for backup 
> > purpose.  DB_CLIENT should be replicated on PGSRV_2. DB_PRODUCT and 
> > DB_ORDER should be replicated on PGSRV_1.
> >
> > So, I would have DB_CLIENT master, DB_PRODUCT slave and 
> DB_ORDER slave 
> > on PGSRV_1 and DB_PRODUCT master, DB_ORDER master and 
> DB_CLIENT slave 
> > on PGSRV_2.
> >
> > So for this replication system, I suppose that I must have 6 nodes, 
> > hence I certainly must have 6 slons too (in the real 
> project, I have 
> > 10 databases and not only 3, so I suppose that would give 
> me 20 slons 
> > running), but how many clusters must/should/could I have?
> 
> If there are three differently-named databases, then you must 
> have three clusters.
> 
> > One possible configuration - and certainly the simplest - 
> is to define 
> > 3 clusters (of 2 nodes each): one for each database to 
> replicate. But 
> > couldn't it be possible to manage less clusters and shouldn't it be 
> > more practical?
> 
> The SET PATH entries have to reference _a database_; they 
> cannot express "get tables from one of these three databases."  
> 
> If you set up one database on each node, and set up three 
> schemas, db_client, db_product, and db_order, then THAT would 
> allow you to cut down to one Slony-I cluster.  That would 
> have the further merit that you could build queries that 
> access data from multiple of those schemas.
> 
> > (I in fact tried to define an unique cluster of 6 nodes, but maybe 
> > have did it wrong because I get the following error at the INIT 
> > CLUSTER command: loading of file /pgsql/share/xxid.v73.sql:
> > PGRES_FATAL_ERROR ERROR:? current transaction is aborted, queries 
> > ignored until end of transaction block)
> 
> Something went wrong; I'm not sure what to point to offhand.
> 
> You might find some relevant info at...
> <http://cbbrowne.dyndns.info:8741/cgi-bin/twiki/view/Sandbox/S
> lonyIAdministration>
> 
> If you find things that are unclear, feel free to add 
> questions to the wiki...
> --
> "cbbrowne","@","ca.afilias.info"
> <http://dev6.int.libertyrms.com/>
> Christopher Browne
> (416) 673-4124 (land)
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://gborg.postgresql.org/pipermail/slony1-general/attachments/20041104/212aa74c/attachment.html