lio bod a écrit :
>  
> You guess right : my purpose is precisly to add a new slave and drop the
> old one.
> Of course, i stop replication process before performing so.
> What are the kind of 'all replication information ' i can loose? So far,
> i can't see what it implies...
>  
> My first step would to retrieve my primary dabase just as before it was
> instrumented by slony. So i can have a 'virgin dump' and restore it in
> any not replicated architecture.
> My second step would be to restart on entirely new replication with a
> new configuration by creating a new cluster and so on.

Then, OK, it would be safe for you, as you really want to clean all nodes from
slony hacking.

>  
> Btw, on my opinion, bare-metal functions are not really industrial. I
> have to supply an integrated solution to a system operator.
> I would rather think of using perl commands just by switching through
> differents slon_tool.conf.
> Any comments on this appoach?

Yes... I never used perl comands because the first time I wanted to use them,
they were buggy, and I din't know perl well enough to be able to repair them.
What's more, these scriptts (as far as I understood them) didn't let me do all
I wanted to (cross-replication, multiple replication from different nodes to a
single spare one...). That's why I used slonik function embedded in shell
scripts I wrote. So now, I have 2 or 3 configuration files and some 10 shell
scripts that allow me to do almost everythng.

That was, in my opinion, the best industrial way of acting.

Regards,
-- 
Stéphane SCHILDKNECHT
Président de PostgreSQLFr
Tél. 09 53 69 97 12
http://www.postgresqlfr.org