Horton

Re: effective_multixact_freeze_max_age issue

Nathan Bossart <nathandbossart@gmail.com>

View thread View on postgr.es
From: Nathan Bossart <nathandbossart@gmail.com>
To: Peter Geoghegan <pg@bowt.ie>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Thomas Munro <thomas.munro@gmail.com>, Robert Haas <robertmhaas@gmail.com>, "Bossart, Nathan" <bossartn@amazon.com>, Jeremy Schneider <schneider@ardentperf.com>
Date: 2022-08-31T03:56:54Z
Lists: pgsql-hackers 
On Tue, Aug 30, 2022 at 05:24:17PM -0700, Peter Geoghegan wrote:
> Attached is v2, which cleans up the structure of
> vacuum_set_xid_limits() a bit more. The overall idea was to improve
> the overall flow/readability of the function by moving the WARNINGs
> into their own "code stanza", just after the logic for establishing
> freeze cutoffs and just before the logic for deciding on
> aggressiveness. That is now the more logical approach (group the
> stanzas by functionality), since we can't sensibly group the code
> based on whether it deals with XIDs or with Multis anymore (not since
> the function was taught to deal with the question of whether caller's
> VACUUM will be aggressive).
> 
> Going to push this in the next day or so.

LGTM

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com

Commits

  1. Derive freeze cutoff from nextXID, not OldestXmin.

  2. Consolidate VACUUM xid cutoff logic.

  3. Fix bug that could try to freeze running multixacts.

  4. Teach autovacuum about multixact member wraparound.