Feedback-ID: ia2694551:Fastmail
Date: Sun, 7 Apr 2024 20:41:41 +0200
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: "Andrey M. Borodin" <x4mmm@yandex-team.ru>
Cc: vignesh C <vignesh21@gmail.com>,
	Andrew Borodin <amborodin86@gmail.com>, i.lazarev@postgrespro.ru,
	Yura Sokolov <y.sokolov@postgrespro.ru>, Andres Freund <andres@anarazel.de>,
	Thomas Munro <thomas.munro@gmail.com>, Gilles Darold <gilles@darold.net>,
	Tomas Vondra <tomas.vondra@enterprisedb.com>,
 Tomas Vondra <tomas.vondra@2ndquadrant.com>,
	Alexander Korotkov <aekorotkov@gmail.com>,
 Anastasia Lubennikova <a.lubennikova@postgrespro.ru>,
	Daniel Gustafsson <daniel@yesql.se>,
 Kyotaro Horiguchi <horikyota.ntt@gmail.com>,
	pgsql-hackers mailing list <pgsql-hackers@postgresql.org>
Subject: Re: MultiXact\SLRU buffers configuration
Message-ID: <202404071841.oxm5ojuahd2q@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <0925F9A9-4D53-4B27-A87E-3D83A757B0E0@yandex-team.ru>
Archived-At: 
 <https://www.postgresql.org/message-id/202404071841.oxm5ojuahd2q%40alvherre.pgsql>
Precedence: bulk

On 2024-Feb-03, Andrey M. Borodin wrote:

> Here's the test draft. This test reliably reproduces sleep on CV when waiting next multixact to be filled into "members" SLRU.
> Cost of having this test:
> 1. We need a new injection point type "wait" (in addition to "error" and "notice"). It cannot be avoided, because we need to sync at least 3 processed to observe condition we want.
> 2. We need new way to declare injection point that can happen inside critical section. I've called it "prepared injection point".
> 
> Complexity of having this test is higher than complexity of CV-sleep patch itself. Do we want it? If so I can produce cleaner version, currently all multixact tests are int injection_points test module.

Well, it would be nice to have *some* test, but as you say it is way
more complex than the thing being tested, and it zooms in on the
functioning of the multixact creation in insane quantum-physics ways ...
to the point that you can no longer trust that multixact works the same
way with the test than without it.  So what I did is manually run other
tests (pgbench) to verify that the corner case in multixact creation is
being handled correctly, and pushed the patch after a few corrections,
in particular so that it would follow the CV sleeping protocol a little
more closely to what the CV documentation suggests, which is not at all
what the patch did.  I also fixed a few comments that had been neglected
and changed the name and description of the CV in the docs.

Now, maybe we can still add the test later, but it needs a rebase.

-- 
Álvaro Herrera        Breisgau, Deutschland  —  https://www.EnterpriseDB.com/
“Cuando no hay humildad las personas se degradan” (A. Christie)