Date: Mon, 24 Jun 2024 19:09:52 -0700
From: Noah Misch <noah@leadboat.com>
To: Peter Geoghegan <pg@bowt.ie>
Cc: Melanie Plageman <melanieplageman@gmail.com>,
	Robert Haas <robertmhaas@gmail.com>,
	Pg Hackers <pgsql-hackers@postgresql.org>,
	Andres Freund <andres@anarazel.de>
Subject: Re: Vacuum ERRORs out considering freezing dead tuples from before
 OldestXmin
Message-ID: <20240625020952.9f.nmisch@google.com>
References: 
 <CAAKRu_bDD7oq9ZwB2OJqub5BovMG6UjEYsoK2LVttadjEqyRGg@mail.gmail.com>
 <CA+TgmoZiArAXM7KwitYWkCmeUQ76Sfg=mgq4_oFo2awWDFdePw@mail.gmail.com>
 <CAH2-Wz=irYFYP5YP2-TajwO3B1maqSSoH-7Ps1aaRv6BHXD7pg@mail.gmail.com>
 <CA+Tgmob4m4TjFCqKFvdR9=Y71fGoH_C-nfPgni6Trt8t3AKOLA@mail.gmail.com>
 <CAAKRu_Y2oTSvhY16EZbLQ6nL3qQXydsfKMXiiCH07KC_GYoRuA@mail.gmail.com>
 <CA+TgmoYzS_bkt_MrNxr5QrXDKfedmh4tStn8UBTTBXqv=3JTew@mail.gmail.com>
 <CAH2-Wzk_nFJOcC6s004bAYGLQ41tPSesTqyBAHcVGxRM97H0Xg@mail.gmail.com>
 <20240625013004.a8.nmisch@google.com>
 <CAH2-Wzn7EqLndKP4pVtxd8QHvTa_Ev_gU=tELz-aaC3Q0yGuTA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: 
 <CAH2-Wzn7EqLndKP4pVtxd8QHvTa_Ev_gU=tELz-aaC3Q0yGuTA@mail.gmail.com>
User-Agent: Mutt/2.2.12 (2023-09-09)
Archived-At: 
 <https://www.postgresql.org/message-id/20240625020952.9f.nmisch%40google.com>
Precedence: bulk

On Mon, Jun 24, 2024 at 09:49:53PM -0400, Peter Geoghegan wrote:
> On Mon, Jun 24, 2024 at 9:30 PM Noah Misch <noah@leadboat.com> wrote:
> > On Mon, Jun 24, 2024 at 03:23:39PM -0400, Melanie Plageman wrote:
> > > Right now, in master, we do use a single horizon when determining what
> > > is pruned -- that from GlobalVisState. OldestXmin is only used for
> > > freezing and full page visibility determinations. Using a different
> > > horizon for pruning by vacuum than freezing is what is causing the
> > > error on master.
> >
> > Agreed, and I think using different sources for pruning and freezing is a
> > recipe for future bugs.  Fundamentally, both are about answering "is
> > snapshot_considers_xid_in_progress(snapshot, xid) false for every snapshot?"
> > That's not to say this thread shall unify the two, but I suspect that's the
> > right long-term direction.
> 
> What does it really mean to unify the two, though?
> 
> If the OldestXmin field was located in struct GlobalVisState (next to
> definitely_needed and maybe_needed), but everything worked in
> essentially the same way as it will with Melanie's patch in place,
> would that count as unifying the two? Why or why not?

To me, no, unification would mean removing the data redundancy.  Relocating
the redundant field and/or code that updates the redundant field certainly
could reduce the risk of bugs, so I'm not opposing everything short of
removing the data redundancy.  I'm just agreeing with the "prefer" from
https://postgr.es/m/CA+TgmoYzS_bkt_MrNxr5QrXDKfedmh4tStn8UBTTBXqv=3JTew@mail.gmail.com