Hi all,

I have recently set up slony replication from an 8.0 DB to an 8.1 DB
with the purpose of postgres upgrade. I've read through the whole
documentation available in the download, and I must say that when
finishing I still didn't know how to install slony... on each machine,
just on the origin, on a third one... not to mention it was not clear
what exact steps are to be done to do the task at hand I had. Sure, I
understood most of the concepts, and what slony can do, but I still
didn't have a step-by-step notion of what to do.

I somehow figured out then that I need to install slony on all DB boxes
and did that. Then installed pgAdmin in the hope it will help creating
the nodes, paths, sets, etc. It didn't - pgAdmin has no task oriented
documentation either and it uses additional concepts which are not
explained at all (like that admin node, I still couldn't figure out how
to use that).

So I went to use the perltools... that looks like the most task oriented
from the whole slony distribution. But there again, the admin
documentation says barely something about how to use them... fortunately
I looked into the README in the perltools directory, and surprise...
contains perfect aid to set up a simple cluster.

> > There is a section with a "task orientation"; we started by constructing
> > a list of "seemingly interesting" tasks on the Wiki
> > <http://slony-wiki.dbitech.ca/index.php/Howto>, and I took the results,
> > extended them, and put that into the admin guide; see the CVS pointer
> > <http://gborg.postgresql.org/cgi-bin/cvsweb.cgi/slony1-engine/doc/adminguide/addthings.sgml?rev=1.17;content-type=text%2Fplain;cvsroot=slony1;only_with_tag=HEAD>

That wiki page is not really visible for somebody who downloads slony
and wants to get on work right away. Whatever documentation is needed
for an admin should be right in the downloaded package, otherwise is
really hard to find (except if you know what to look for).

> > The point of what I could write is that it is from the dba on the street
> perpective, rather than the slony task perspective.  You are working, as you
> should be, on a slony task level.  The higher level would be more like:  Why do you want
> a replica?  A recommended environment set up technique which uses perltools. 
> How to fix your db so it has PKs and UI and why. What is failover (as opposed
> to move set)?  These are off the top of my head. But I go over each step carefully
> with every client that I've set up slony.  
> 
For somebody who installs slony himself, the install instructions are
not that useful... they go into great detail of how to install, but the
big picture is missing. I mean slony is to be installed on multiple
machines, and if you have multiple postgres versions then you might need
to compile it multiple times, that's not documented at all... some kind
of deploy topology is needed. And mentioning that installing slony and
set it up are different tasks... which is not so obvious for a new-by.

> At some point, many of my clients get down to the level of the slony documentation
> but certainly not all of them.  But in order to get there, the non-trivial
> understanding and set up needs to be covered in a language that they understand.
> 
> So this will not be in the same style at all as what is there already.  But I like
> the idea of being able to cross reference.  Perhaps a large addendum tutorial document
> would be appropriate instead of changing the existing work?
> 
> BTW I have accidently found myself on the wiki for the howto several times,  Shouldn't
> the link be more prominent?  I would need cvs privileges in order to change
> the sgml, but it may be possible to funnel that through fetter or some other committer.
> 
> I'd like to hear concerns about too many documentation branches as well as 
> real life problems that people have run into when setting up and administering slony.

For me the following would have been useful:
  - high level description of the concepts (cluster, node, set, etc.) -
this is adequate as it is in the docs now;
  - concise installation description, and a deployment topology aid;
  - task oriented setup instructions, the ones in the perltools README
file helped me the best;

Otherwise, it looks like slony works fine for us... Thanks for another
piece of good software :-)

Cheers,
Csaba.