2007/10/30, Thomas Hallgren <thomas at tada.se>:
> Hi,
> I'm trying to wrap my head around Slony-I and how to use it from an
> application server like JBoss with Hibernate. I'm not clear on the
> relationship between actual connections established by a client and
> nodes in the cluster.
>
> Is it true that any connection that is established is either connected
> to a slave or to the master (origin) for a specific replication set?
> I.e. there's no node in between that will redirect updates to the master
> for the replication set in question?
>
> I'm trying to figure out how to do a controlled switchover and the
> implications of doing that. My appservers can work with two different
> connections if needed but it will be somewhat cumbersome if they need to
> change their settings when I do a switchover. Is there a way to do
> switchover and maintain the connection settings (i.e. host, database,
> credentials)? Ideally, the connections should survive the switch.
>
> I'm also looking into automatic switchover in case the master fails for
> some reason. The Heartbeat product seems to be useful for this. Are you
> aware of any examples on how to do this? Any pointer or hint would be
> useful.


It seems you need pgpool-II + slony + heartbeat(or similar) setup

I heard gossips that there are actually a few people in the world who
managed to do this :)

No single ready made HOWTO exists I know of.

google for pgpool+slony+heartbeat.
there are some interesting documents, including presentations from PGCon.

HTH

-- 
Filip Rembiałkowski