Horton

Re: recovering from "found xmin ... from before relfrozenxid ..."

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Robert Haas <robertmhaas@gmail.com>
Cc: Ashutosh Sharma <ashu.coek88@gmail.com>, Mark Dilger <mark.dilger@enterprisedb.com>, Masahiko Sawada <masahiko.sawada@2ndquadrant.com>, "Andrey M. Borodin" <x4mmm@yandex-team.ru>, MBeena Emerson <mbeena.emerson@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Andres Freund <andres@anarazel.de>, Peter Eisentraut <peter.eisentraut@2ndquadrant.com>, Stephen Frost <sfrost@snowman.net>
Date: 2020-09-15T19:54:56Z
Lists: pgsql-hackers 
Robert Haas <robertmhaas@gmail.com> writes:
> On Mon, Sep 14, 2020 at 6:26 AM Ashutosh Sharma <ashu.coek88@gmail.com> wrote:
>> Thanks for reporting. I'm able to reproduce the issue by creating some
>> delay just before "-- now create an unused line pointer" and use the
>> delay to start a new session either with repeatable read or
>> serializable transaction isolation level and run some query on the
>> test table. To fix this, as you suggested I've converted the test
>> table to the temp table. Attached is the patch with the changes.
>> Please have a look and let me know about any concerns.

> Tom, do you have any concerns about this fix?

It seems OK as far as it goes.  Two thoughts:

* Do we need a comment in the test pointing out that the table must be
temp to ensure that we get stable vacuum results?  Or will the commit
log message be enough documentation?

* Should any of the other tables in the test be converted to temp?
I see that the other test cases are kluging around related issues
by dint of never committing their tables at all.  It's not clear
to me how badly those test cases have been distorted by that, or
whether it means they're testing less-than-typical situations.

Anyway, if you're satisfied with leaving the other cases as-is,
I have no objections.

			regards, tom lane

Commits

  1. Fix wrong data table horizon computation during backend startup.

  2. Centralize horizon determination for temp tables, fixing bug due to skew.

  3. pg_surgery: Try to stabilize regression tests.

  4. New contrib module, pg_surgery, with heap surgery functions.

  5. Set cutoff xmin more aggressively when vacuuming a temporary table.

  6. snapshot scalability: Don't compute global horizons while building snapshots.

  7. Introduce vacuum errcontext to display additional information.