John Sidney-Woollett <johnsw at wardbrook.com> writes:
> I have read lots of slony docs and examples but am unclear about this issue.
>
> If I have a shell script which defines the cluster, two nodes (on two
> separate hosts), the sets, paths and listen and subscribe
> information...
>
> Do I just need to run this script on the master server only, or do I
> run the same or a cust down version of the script on the slave server
> too?
>
> In other words, does the act of running the slonik config script for
> my new cluster on the first server propogate all the replication
> config information required to all nodes automatically (by
> connecting to each server in the cluster in turn)?

You're likely causing yourself confusion any time you use the terms
"master" or "slave," as they don't really properly describe what is
going on.

There's several parts to Slony-I configuration:

0.  Each script contains a preamble indicating how slonik will connect
    to the nodes.  That's repeated "gratuitously."

1.  You define the nodes, which kind of covers _all_ of the cluster.

2.  You define a set, and indicate what tables are in the set, as well
    as what node is the origin.  This controls what each set looks like
    across the cluster.

3.  Finally, you tell some node to subscribe to the set through some
    provider.

    This leads to the slon process for the subscriber rummaging around
    on that provider to look at what tables and sequences are in the
    set.

Thus, when you add a subscriber, all you need do is to say "subscribe
Node A to that set," which is what's in part #3.  Node A _draws in_
the information that it needs about the replication set.
-- 
"cbbrowne","@","ca.afilias.info"
<http://cbbrowne.com/info/slony.html>
Christopher Browne
(416) 673-4124 (land)