Re: PITR and base + full backups

On Tuesday 16 September 2008, "Joey K." <pguser(at)gmail(dot)com> wrote:
> Hello,
>
> Just to be sure of our backups we plan to do a base + full backups (yes,
> we are overly paranoid)
>
> (1) (`date`)
>
> (2) perform hot rsync first (while the database is running)
> $ rsync -avr pgdata /backup/`date`/
>
> (3) stop pg
>
> (4) perform cold rsync
> $ rsync -avr --delete pgdata /backup/`date`/
>
> (5) start pg
>
> (6) pg_stop_backup()
>
> This didn't work and not sure if this is supposed to work ;-)
>
> Or should I stick to just plain PITR?

If you can live with the downtime to do that, you don't need steps 1 or 6,
and eliminating them will probably fix your problem. I imagine stopping the
database between pg_start_backup and pg_stop_backup caused whatever problem
you experienced.

Plain PITR is better, though, if you're actually archiving the WAL logs, as
your possibly recovery points will be as current as your last archived log.

--
Alan