On Wed, Aug 27, 2008 at 01:37:22AM +0800, Phoenix Kiula wrote:

> A simple scenario -- typical one! -- would go a long way in making
                                     ^^^^^^^^^

There _is_ a typical scenario in there.  There's a significant
tutorial at the beginning.  Did you do it?

> 2. Install new instance of PG. Take a typical PG install. Include
> thoughts on installing in a NEW DIRECTORY so that it doesn't conflict
> with existing PG.

This is perfectly clear in the Postgres documentation.  Slony is an
add-on to Postgres.  Not a replacement.  You need to have two Postgres
instances.  Why should the Slony docs reproduce how to do that?  If
you need Slony, you already have some Postgresses around, it's
assumed.  I don't think that assumption strains credulity.
 
> 3. Do the master/slave stuff. I find it hard to believe that there is
> NO default situation. There has to be a default config, right? Even if

No, there doesn't, and there isn't, and that's exactly what everyone
is trying to tell you.  Even the term "master/slave" is wrong.  This
is why I said you need to go read the manual, long though it may be.
Alternatively, as I suggested in another thread, you might want to
look at Londiste, which has many fewer features precisely because
those people found Slony's surfeit of features too complicated.  As,
apparently, you do.  I understand that: I often find it too
complicated, too, and I was around from the moment the thing was
imagined.  

> The "concepts" page tells me nothing. I have read it, and the ENTIRE
> slony documentation, plus some google pages here and there. Not a
> single thing that explains how to do stuff.

It is entirely possible that you have run your eyes over the ENTIRE
Slony documentation, but I find it impossible to believe you "read" it
(in the sense of having understood all the words), but be able to
claim it contains "not a single thing that explains how to do stuff".
I have some issues with the way the documents are, too.  But how to
set up and run your first slony instance certainly is in there.  I
have personally run through the steps, and I am sure they work.  That
manual does explain how to do stuff.

It doesn't tell you what to do for your particular case, because it's
a general purpose tool.  You have the tools to build a house, and all
you want is to fix your faucet.  I get that.  The point we've been
trying to make is that you have two choices:

1.  Learn how to do enough home repair with the tools you have that
you can fix the faucet.

2.  Call a plumber.  (I've also noted that I am in fact such a
plumber, and if you want such help, you can call Command Prompt or
other companies who specialise in supporting cases like yours.  Somone
else pointed you to a community-supplied list of such companies.  Take
your pick.)
 
> looking for regular replication, but I was forwarded from the
> PG-GENERAL list to come here and ask how to use Slony to perform the
> anal experience that is the PG upgrade. I am not impressed so far!)

I think if you'd continued to read in that thread, you'd have noticed
that I actually suggested that learning Slony might be a big deal for
someone like you, who was surprised that dump and restore was the
standard Postgres upgrade approach.  (I confess that I find it more
than a little depressing that people who can't afford 8 hours of
outage don't know about these problems before they run into them in
production, but I guess we get the Internet we deserve.)

A

-- 
Andrew Sullivan
ajs at commandprompt.com
+1 503 667 4564 x104
http://www.commandprompt.com/