Message-ID: <16d61d864feda279cfa6bcb4ea4d50c2b4fa3a7f.camel@cybertec.at>
Subject: Re: Tooling for per table autovacuum tuning
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Joseph Hammerman <joe.hammerman@datadoghq.com>, Wolfgang Wilhelm
	 <wolfgang20121964@yahoo.de>
Cc: "pgsql-admin@lists.postgresql.org" <pgsql-admin@lists.postgresql.org>
Date: Mon, 13 Mar 2023 06:58:44 +0100
In-Reply-To: 
 <CAHs7QM_dDN3M=G0vEEi7V_wZaWmtJO5OK6ruh7bBfQ4BNo4QxQ@mail.gmail.com>
References: 
	<CAHs7QM_scGEh8uhsCUoc_BEimyYsUTVZxpE10rUiqnShd8nTkQ@mail.gmail.com>
	 <5c0a6ab9ba801e0fa7c2217e49ce804837c64cf1.camel@cybertec.at>
	 <14072753.2216234.1678640212317@mail.yahoo.com>
	 <CAHs7QM_dDN3M=G0vEEi7V_wZaWmtJO5OK6ruh7bBfQ4BNo4QxQ@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.46.4 (3.46.4-1.fc37) 
MIME-Version: 1.0
Archived-At: 
 <https://www.postgresql.org/message-id/16d61d864feda279cfa6bcb4ea4d50c2b4fa3a7f.camel%40cybertec.at>
Precedence: bulk

On Sun, 2023-03-12 at 13:34 -0700, Joseph Hammerman wrote:
> What I am imagining is per table tuning that buckets the tables based on =
their relative sizes.
>=20
> Something like:
>=20
> Up to 1Gb - Small
> Up to 4Gb - Medium
> Up to 8Gb - L
> Bigger - XL
>=20
> And an accordant autovacuum_scale_factor associated with each size.
>=20
> The motivation for this is to make sure large tables get regularly vacuum=
ed.

I understand.

There is no such option currently.  Perhaps you can use "autovacuum_vacuum_=
threshold"
for something like that: if you set it to a certain (high) value and set "a=
utovacuum_vacuum_scale_factor"
to 0, autovacuum is triggered based on the absolute number of dead tuples.

But I would say that the standard configuration makes sense in this case: n=
ormally,
large tables don't need to be vacuumed that often, and vacuum on large tabl=
es is
nore expensive too.

Yours,
Laurenz Albe