Horton

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

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Amit Kapila <amit.kapila16@gmail.com>, Robert Haas <robertmhaas@gmail.com>, 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>, Peter Eisentraut <peter.eisentraut@2ndquadrant.com>, Stephen Frost <sfrost@snowman.net>
Date: 2020-10-28T03:51:10Z
Lists: pgsql-hackers

Attachments

Hi,

On 2020-10-15 01:37:35 -0700, Andres Freund wrote:
> Attached is a *prototype* implemention of this concept, which clearly is
> lacking some comment work (and is intentionally lacking some
> re-indentation).
> 
> I described my thoughts about how to limit the horizons for temp tables in
> https://www.postgresql.org/message-id/20201014203103.72oke6hqywcyhx7s%40alap3.anarazel.de

Attached is an updated version of this patch. Quite a bit of polish,
added removal of the isTopLevel arguments added a7212be8b9e that are now
unnecessary, and changed the initialization of the temp table horizons
to be latestCompletedXid + 1 instead of just latestCompletedXid when no
xid is assigned.


> Besides comments this probably mainly needs a bit more tests around temp
> table vacuuming. Should have at least an isolation test that verifies
> that temp table rows can be a) vacuumed b) pruned away in the presence
> of other sessions with xids.

I added an isolationtester test for this. It verifies that dead rows in
temp tables get vacuumed and pruned despite concurrent sessions having
older snapshots. It does so by forcing an IOS and checking the number of
heap fetches reported by EXPLAIN. I also added a companion test for
permanent relations, ensuring that such rows do not get removed.


Any comments? Otherwise I'll push that patch tomorrow.

Greetings,

Andres Freund

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.