> > To clarify:
> > I'd expect a cluster to be workable, if I
> > - disable VACUUM until backup completed
> > - issue CHECKPOINT
> > - backup clog (CHECKPOINT and backup clog are the "backup checkpoint")
> > - backup all datafiles (which include at least all completed transaction
> > data at checkpoint time)
> > and then
> > - restore datafiles and clog
> > - bring up pgsql.
>
> Why is that a useful approach? You might as well shut down the
> postmaster and do a cold filesystem backup, because you are depending on
> the data files (including clog) not to change after the checkpoint. You
> cannot make such an assumption in a running database.
I think there is a misunderstanding here.
What I think is possible is the following (continuous backup of WAL assumed):
- disable VACUUM
- issue CHECKPOINT "C1"
- backup all files
- reenable VACUUM
- restore files
- adapt pg_control (checkpoint "C1")
- recover WAL until at least end of backup
The db is inconsistent until you recovered all WAL (PITR) that accumulated during
file backup.
I am not sure about clog, isn't clog logged in xlog ?
Andreas