Chris Hoover wrote:

>I have a questions about slony1.
>
>1.  Are there any rpm's available for it?  I'm running redhat 2.1 in
>production, and looking at redhat 4.0 for our next upgrade cycle.  We
>are really trying to get away from compiling "system" level programs,
>and using rpms for them.  If we were to use slony1, we would really
>prefer to install from a standard rpm and not have to worry about
>installing the compilers and other development rpms on our production
>db servers.
>  
>
Some hard work went into this very matter in the last week; there are
now RPM .spec files that can presumably be used with some versions of
RHAT's distributions.  I wouldn't be surprised to see packages available
some time soon.

>2.  Can slony1 replicate the data between different version of
>Postgresql?  We are currently running 7.3.4 in production, but are
>looking at 8.0 for during the next upgrade cycle.  Would it be
>possible to use slony1 to replicate between the 2 servers (assuming
>identical schema's) and then switch our users to the 8.0 server when
>ready?
>  
>
This is one of the conscious "use cases" for Slony-I.

You're in luck:  7.3.4 is newer than 7.3.3, and thus falls into the set
of versions supported.

Unfortunately, you're not terribly likely to see RPMs for 7.3.4;
distribution makers aren't likely to go to the trouble of building out
versions for old PG versions.

>3.  We have each of our customers in a different database for security
>reasons.  How much overhead is there in running slony1?  Most db
>servers have 20 or less db's on them.  However we do have atleast 1
>server with 150+ db's on it.
>  
>
Ah, I think you may find it somewhat expensive for such cases :-(.

For each distinct database on each node, you'll need a slon daemon that
opens up at least 2 connections.  (One to the node it's 'feeding', and
one to its provider).

If you are running 3 nodes (e.g. - 'master' + 2 'slaves') for each
database, then you'll have rather a lot of database connections going
into the various systems.

That server with 150 DBs would lead to 150 slon processes, each
connecting to that host, as well as to at least one other, meaning you
would have at least 300 DB connections kicking around for this.

I daresay that could get expensive :-(.

If some of them are only barely used (which is pretty likely), you'll
find that Slony-I is a lot more costly than the normal activity :-(.

One thing that is VERY important to note is that you had better be
prepared to apply tight reins to transaction behaviour on these
systems.  A buggy client that leaves idle connections open for days will
have adverse effects on performance for EVERY ONE of the databases on
that host with 150+ DBs.  Vacuums won't take effect, and Slony-I really
needs for its tables to be vacuumed regularly.