Horton

Re: Non-replayable WAL records through overflows and >MaxAllocSize lengths

Robert Haas <robertmhaas@gmail.com>

View thread View on postgr.es
From: Robert Haas <robertmhaas@gmail.com>
To: Heikki Linnakangas <hlinnaka@iki.fi>
Cc: Matthias van de Meent <boekewurm+postgres@gmail.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2022-03-11T21:12:30Z
Lists: pgsql-hackers 
On Fri, Mar 11, 2022 at 3:42 PM Heikki Linnakangas <hlinnaka@iki.fi> wrote:
> Have you been able to create a test case for that? The largest record I
> can think of is a commit record with a huge number of subtransactions,
> dropped relations, and shared inval messages. I'm not sure if you can
> overflow a uint32 with that, but exceeding MaxAllocSize seems possible.

I believe that wal_level=logical can generate very large update and
delete records, especially with REPLICA IDENTITY FULL.

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

Commits

  1. Add more protections in WAL record APIs against overflows

  2. Add overflow protection for block-related data in WAL records

  3. Change internal RelFileNode references to RelFileNumber or RelFileLocator.

  4. Revamp the WAL record format.