Alan Hodgson <ahodgson at simkin.ca> writes:
> On Wed, Sep 01, 2004 at 04:25:00PM -0400, Christopher Browne wrote:
>> (I think users would be doing something pretty stupid to have _wildly_
>> different arrangements for different sets, but I still have to support
>> it...)
>
> Actually I've thought of some interesting things to do with replication.
>
> Currently I replicate a subset of my tables to each member of a web
> cluster (that need fast and reliable read-only access to the data).
> I might add a full backup server, which would need to subscribe to
> all the data.  I'm also considering running PowerDNS, driven from
> pgsql, in which case I might replicate a different subset of the
> tables to the DNS servers, again for fast, reliable read-only access
> to the data.
>
> I think all of these together would count as fairly different
> arrangements for different sets.
>
> Just a data point.

If the "arrangements" of different sets vary sufficiently, it might
also make sense to have multiple Slony-I clusters, particularly if
different sets are talking to different databases.

I still would think that _really_ wacky configurations would be pretty
dumb, as they make it easy to "fat finger" things and accidently trash
something.  (That's not too hard to accomplish, even _without_ doing
anything unusual , unfortunately:-(.)

In any case, we could really use an article on how people have
configured clusters, particularly for cases more unusual than the mere
"I set up an origin for my application, to which a couple of nodes
subscribe."  If you could do a writeup of how you set things up, that
could illuminate some of the "dark spaces" for others.

We have some of the early "theoretical" documentation on system
design, but could REALLY use more material along the lines of

  "I want to use this Slony-I system; how should I configure it?"
-- 
let name="cbbrowne" and tld="ca.afilias.info" in name ^ "@" ^ tld;;
<http://dev6.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)