On 7/25/07, Geoffrey <lists@serioustechnology.com> wrote:
> We plan to upgrade from our current 7.4.17 Postgresql to 8.2.? in the
> near future. We also plan to implement a replication solution, most
> likely Slony.
>
> We know that the upgrade to 8.2.? will require some code changes.
>
> Question is, is there a particular order in which we should perform
> these changes that would require the least amount of additional work?
>
> That is, would it be more efficient to move to 8.2.? and then implement
> our Slony replication, or would there be less effort in implementing the
> Slony solution and then upgrading to 8.2.? ?
absolutely I would suggest to upgrade to 8.2 unless you already were
on 8.1. note this is not really having anything to do with
slony...it's just that for a 24/7 environment 8.2 is so much easier to
manage with autovacuum, pitr, and better performance all around. If
you were extremely paranoid 8.1 would also be on the table since it's
more mature and is the version provided by most of the enterprise
distros, and there are relatively few killer features on 8.2...IMO,
the most relevant to 24x7 env would be the improved autovac accounting
so you can more easily see what is going on. #2 feature for 24x7 on
8.2 is create index concurrently.
IIRC you can use slony to move the data from the 7.4 to the 8.2
machine and then minimize your downtime (this is one of the nicest
things about slony). In that sense you may want to look at setting up
slony first.
merlin