Horton

Re: pg15b2: large objects lost on upgrade

Robert Haas <robertmhaas@gmail.com>

View thread
From: Robert Haas <robertmhaas@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: "Jonathan S. Katz" <jkatz@postgresql.org>, Peter Geoghegan <pg@bowt.ie>, Andres Freund <andres@anarazel.de>, Noah Misch <noah@leadboat.com>, Andrew Dunstan <andrew@dunslane.net>, Bruce Momjian <bruce@momjian.us>, Michael Paquier <michael@paquier.xyz>, Justin Pryzby <pryzby@telsasoft.com>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>, Shruthi Gowda <gowdashru@gmail.com>
Date: 2022-08-04T16:43:49Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Revert recent changes to 002_pg_upgrade.pl.

  2. Revise test case added in 43746996399541ecb5c7b188725a5f097c15ceae.

  3. Use TRUNCATE to preserve relfilenode for pg_largeobject + index.

  4. Preserve relfilenode of pg_largeobject and its index across pg_upgrade.

  5. Have VACUUM warn on relfrozenxid "in the future".

  6. Set relfrozenxid to oldest extant XID seen by VACUUM.

  7. pg_upgrade: Preserve relfilenodes and tablespace OIDs.

  8. Perform a lot more sanity checks when freezing tuples.

On Thu, Aug 4, 2022 at 10:26 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
> > 100 << 2^32, so it's not terrible, but I'm honestly coming around to
> > the view that we ought to just nuke this test case.
>
> I'd hesitated to suggest that, but I think that's a fine solution.
> Especially since we can always put it back in later if we think
> of a more robust way.

IMHO it's 100% clear how to make it robust. If you want to check that
two values are the same, you can't let one of them be overwritten by
an unrelated event in the middle of the check. There are many specific
things we could do here, a few of which I proposed in my previous
email, but they all boil down to "don't let autovacuum screw up the
results".

But if you don't want to do that, and you also don't want to have
random failures, the only alternatives are weakening the check and
removing the test. It's kind of hard to say which is better, but I'm
inclined to think that if we just weaken the test we're going to think
we've got coverage for this kind of problem when we really don't.

-- 
Robert Haas
EDB: http://www.enterprisedb.com