Thomas F.O'Connell <tfo at sitening.com> writes:
> In the README and in the docs, there are basically a pair of scripts:
> one for setup and one for subscription. What is unclear (to me) is on
> which nodes the scripts themselves need to be run.
>
> Once the daemon is running on both a master and a slave node, does it
> matter which node actually runs the setup and subscription scripts?
>
> Thanks!

I think you're confusing here:
 - Where do I run slonik, which talks to database instances, and
   tells them how to control what's going on?

   Slonik should be able to run anywhere that can talk to one or more
   of the DBs; it initiates connections to the databases based on the
   conninfo strings you pass in.

 - Where do I run slon, which is the daemon that controls a particular
   replication instance?

   The initial assumption was that that should run on the DB server.

After some experimenting this week, it turns out that it doesn't
matter TOO much where slon instances run, as long as origins and
subscribers are on a local network.

But suppose a subscriber database is quite distant, such as on a WAN
somewhere.  (e.g. - Connected to its "parents" via an ADSL connection)
In THAT case, you want the slon daemon for that database to sit on the
same box as the database server that it's to feed, because it does
particularly a lot of communicating with that database.

By the way, I'll Real Soon Now be putting into CVS a set of Perl
scripts that represent a somewhat more sophisticated set of
functionality:

 - They support having a whole "tree" of Slony-I nodes, not just
   a 'master' and a 'slave.'
 - There are separate scripts for the many actions you might want
   to perform.  Initializing, re-initializing, adding sets, 
   adding subscribers, and such.
-- 
(reverse (concatenate 'string "ofni.sailifa.ac" "@" "enworbbc"))
<http://dev6.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)