Content-Type: text/plain;
	charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3696.120.41.1.8\))
Subject: Re: MultiXact\SLRU buffers configuration
From: "Andrey M. Borodin" <x4mmm@yandex-team.ru>
In-Reply-To: <20240229.105944.773077833314087742.horikyota.ntt@gmail.com>
Date: Sat, 6 Apr 2024 16:24:12 +0500
Cc: Alvaro Herrera <alvherre@alvh.no-ip.org>,
 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.net,
 tomas.vondra@enterprisedb.com,
 tomas.vondra@2ndquadrant.com,
 aekorotkov@gmail.com,
 a.lubennikova@postgrespro.ru,
 daniel@yesql.se,
 pgsql-hackers@postgresql.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <8401378F-62C2-4E5D-A2AC-F8C8B7D2FF5B@yandex-team.ru>
References: <202401281249.u6ham26mbjav@alvherre.pgsql>
 <AE4A1D43-BB1C-45AB-A964-8455B2E60F8F@yandex-team.ru>
 <0925F9A9-4D53-4B27-A87E-3D83A757B0E0@yandex-team.ru>
 <20240229.105944.773077833314087742.horikyota.ntt@gmail.com>
To: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Archived-At: <https://www.postgresql.org/message-id/8401378F-62C2-4E5D-A2AC-F8C8B7D2FF5B%40yandex-team.ru>
Precedence: bulk


> On 29 Feb 2024, at 06:59, Kyotaro Horiguchi <horikyota.ntt@gmail.com> =
wrote:
>=20
> At Sat, 3 Feb 2024 22:32:45 +0500, "Andrey M. Borodin" =
<x4mmm@yandex-team.ru> wrote in=20
>> Here's the test draft. This test reliably reproduces sleep on CV when =
waiting next multixact to be filled into "members" SLRU.
>=20
> By the way, I raised a question about using multiple CVs
> simultaneously [1]. That is, I suspect that the current CV
> implementation doesn't allow us to use multiple condition variables at
> the same time, because all CVs use the same PCPROC member cvWaitLink
> to accommodate different waiter sets. If this assumption is correct,
> we should resolve the issue before spreading more uses of CVs.

Alvaro, Kyotaro, what's our plan for this?
It seems to late to deal with this pg_usleep(1000L) for PG17.
I propose following course of action
1. Close this long-standing CF item
2. Start new thread with CV-sleep patch aimed at PG18
3. Create new entry in July CF

What do you think?


Best regards, Andrey Borodin.=