Horton

Re: Doc fix of aggressive vacuum threshold for multixact members storage

Bertrand Drouvot <bertranddrouvot.pg@gmail.com>

View thread View on postgr.es
From: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
To: Sami Imseih <samimseih@gmail.com>
Cc: Alex Friedman <alexf01@gmail.com>, pgsql-hackers@lists.postgresql.org
Date: 2025-02-21T09:43:43Z
Lists: pgsql-hackers 
Hi,

On Tue, Feb 04, 2025 at 01:23:28PM -0600, Sami Imseih wrote:
> I confirmed the 20GB value as is described here [1].
> 8k page can hold 409 member groups and each
> member group can hold 4 members, thus
> (2^32/(409 *4))*8192 = 20GB.
> 

Thanks for the patch!

I do agree that that's a bug in the doc.

Indeed, while I do agree that MULTIXACT_MEMBER_SAFE_THRESHOLD can be seen
as "2G" (not 2GB...) then it's compared in MultiXactMemberFreezeThreshold() to
"members" which is the number of members: So, the member's size has to be taken
into account.

Maybe we could also add a comment in multixact.c to update the doc accordingly if
the computation changes? (I think that will be easy to miss).

Regards,

-- 
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com

Commits

  1. Doc: correct aggressive vacuum threshold for multixact members storage

  2. docs: clarify new aggressive vacuum mode for multi-xacts

  3. Increase threshold for multixact member emergency autovac to 50%.