On Wed, May 13, 2020 at 1:54 AM Robert Haas <robertmhaas@gmail.com> wrote:
>
> On Tue, May 12, 2020 at 2:23 AM Amit Kapila <amit.kapila16@gmail.com> wrote:
> > While looking at this, I noticed that caller (perform_base_backup) of
> > do_pg_start_backup, sets the backup phase as
> > PROGRESS_BASEBACKUP_PHASE_WAIT_CHECKPOINT whereas, in
> > do_pg_start_backup, we do collect the information about all
> > tablespaces after the checkpoint. I am not sure if it is long enough
> > that we consider having a separate phase for it. Without your patch,
> > it was covered under PROGRESS_BASEBACKUP_PHASE_ESTIMATE_BACKUP_SIZE
> > phase which doesn't appear to be a bad idea.
>
> Maybe I'm confused here, but I think the size estimation still *is*
> covered under PROGRESS_BASEBACKUP_PHASE_ESTIMATE_BACKUP_SIZE. It's
> just that now that happens a bit later.
>
There is no problem with this part.
> I'm assuming that listing the
> tablespaces is pretty cheap, but sizing them is expensive, as you'd
> have to iterate over all the files and stat() each one.
>
I was trying to say that tablespace listing will happen under
PROGRESS_BASEBACKUP_PHASE_WAIT_CHECKPOINT phase which could be a
problem if it is a costly operation but as you said it is pretty cheap
so I think we don't need to bother about that.
Apart from the above point which I think we don't need to bother, both
your patches look good to me.
--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com