On Wed, Apr 14, 2010 at 1:42 PM, Kenneth Marshall
<ktm@rice.edu> wrote:
On Wed, Apr 14, 2010 at 01:29:23PM -0400, Lewis Kapell wrote:
> In the documentation under the heading "Migration Between Releases" we
> read:
>
> <<It is also possible to use replication methods, such as Slony, to create
> a slave server with the updated version of PostgreSQL. The slave can be on
> the same computer or a different computer. [...] Such a switch-over results
> in only several seconds of downtime for an upgrade. >>
>
> In the section "Warm Standby Servers for High Availability" it says:
>
> << For testing purposes, it is possible to run both primary and standby
> servers on the same system. >>
>
> Although this section does not use the term Point-In-Time Recovery, I
> understand this is what it refers to. Could the WAL/PITR method then be
> used to upgrade from 8.3 to 8.4 on the same server and avoid having the
> significant downtime of dump/restore?
>
> If so, I understand the two servers would have to run on different ports.
> Are there any other issues/traps to be aware of?
>
> --
>
> Thank you,
>
> Lewis Kapell
> Computer Operations
> Seton Home Study School
>
I believe that the server versions must be the same in PITR so
you cannot use it to upgrade.
Exactly, remember, one of the most important steps in PITR is the 'base' backup. Your PITR slaves uses the same data files as its starting point as the master. Since those aren't upgradeable (except in a few cases with pg_migrator) you're really out. Even if you did get pg_migrator to upgrade the base files, I wouldn't use them in a PITR setup though. I'm not 100% sure if it changed or not, but I would assume that the WAL record format changed from version to version.
--Scott
Cheers,
Ken