Re: Should pg_dumpall dump ALTER SYSTEM settings?

At Thu, 07 Apr 2022 12:38:43 +0200, Laurenz Albe <laurenz.albe@cybertec.at> wrote in 
> On Wed, 2022-04-06 at 21:39 -0400, Robert Haas wrote:
> > On Wed, Apr 6, 2022 at 2:26 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > > Thoughts?
> > 
> > I'm a little bit skeptical about this proposal, mostly because it
> > seems like it has the end result that values that are configured in
> > postgresql.conf and postgresql.auto.conf end up being handled
> > differently: one file has to be copied by hand, while the other file's
> > contents are propagated forward to the new version by pg_dump. I don't
> > think that's what people are going to be expecting...
> 
> "postgresql.auto.conf" is an implementation detail, and I would expect
> most users to distinguish between "parameters set in postgresql.conf"
> and "parameters set via the SQL statement ALTER SYSTEM".
> If that is the way you look at things, then it seems natural for the
> latter to be included in a dump, but not the former.
> 
> As another case in point, the Ubuntu/Debian packages split up the data
> directory so that the config files are under /etc, while the rest of
> the data directory is under /var/lib.  "postgresql.auto.conf" is *not*
> in /etc, but in /var/lib there.  So a user of these distributions would
> naturally think that the config files in /etc need to be handled manually,
> but "postgresql.auto.conf" need not.
> 
> I am +1 on Tom's idea.

I'm -0.2 if it is the default/implicit behavior. postgresql.conf and
ALTER SYSTEM SET works on the same set of settings.  If we include
.auto's settings in a dump data, it overrides the intentional changes
in postgresql.conf.  I see it a bit surprising.

I'm +-0 if it is a optional behavior.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center