Content-Type: text/plain;
	charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3864.400.21\))
Subject: Re: doc: add note that wal_level=logical doesn't set up logical
 replication in itself
From: Chao Li <li.evan.chao@gmail.com>
In-Reply-To: <093b33a0-6862-4b78-b778-d54e29175c7e@app.fastmail.com>
Date: Thu, 5 Mar 2026 10:20:22 +0800
Cc: Florents Tselai <florents.tselai@gmail.com>,
 Masahiko Sawada <sawada.mshk@gmail.com>,
 pgsql-hackers <pgsql-hackers@lists.postgresql.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <EA37BD89-FD8A-4A8B-8C21-0CAC8636B087@gmail.com>
References: 
 <CA+v5N42vKtY17653eBHFx+Cw06E2fSmPrXmX2qKYG23=4Ycd7A@mail.gmail.com>
 <CAD21AoAFJjLY4GF1+MErybkTcbvFvzUZAV_Nr0wviE-Z7fNcMw@mail.gmail.com>
 <32478450-80EC-4C55-9C80-35BEFB62060E@gmail.com>
 <CA+v5N40pDJpsd4r_5Rcv=aMy_6OxBqc-nfX_3bbLfASWmhbE4A@mail.gmail.com>
 <093b33a0-6862-4b78-b778-d54e29175c7e@app.fastmail.com>
To: Euler Taveira <euler@eulerto.com>
Archived-At: 
 <https://www.postgresql.org/message-id/EA37BD89-FD8A-4A8B-8C21-0CAC8636B087%40gmail.com>
Precedence: bulk


> On Mar 4, 2026, at 22:11, Euler Taveira <euler@eulerto.com> wrote:
>=20
> On Wed, Mar 4, 2026, at 8:34 AM, Florents Tselai wrote:
>>=20
>> +        <literal>logical</literal> adds information necessary to =
support
>> +        <link linkend=3D"logical-replication">logical =
replication</link>, which
>> +        relies on logical decoding. Each level includes the =
information logged
>> +        at all lower levels.
>>=20
>=20
> I'm not opposed to this idea. However, I would like to point out that
>=20
> 1) "replication" is mentioned for _replica_ which can introduce some =
confusion
> if we add "logical replication". Maybe it is a good idea to change the =
sentence
> below from "replication" to "physical replication".
>=20
>> The default value is replica, which writes enough data to support WAL
>> archiving and replication

I agree with the point. But I think =E2=80=9Cstreaming replication=E2=80=9D=
 is more common: =E2=80=9Cto support WAL archiving and streaming =
replication=E2=80=9D.

>=20
> 2) There is no link to the sentence above. If you add a link to the =
logical
> replication section, I would suggest to add a link to the other cases =
(WAL
> archiving, physical replication) too.
>=20

I have no objection on updating other docs, but I don=E2=80=99t consider =
that is a must for the current patch. As Tom mentioned in the other =
thread, there have been already a lot of inconsistencies, we cannot =
expect all docs written from a single pen.

> 3) If we are using "logical replication", should we mention "logical =
decoding"?
> The Logical Replication chapter mentions the use of logical decoding =
in the
> Architecture section.
>=20

IMO, we should. Because the option actually enables =E2=80=9Clogical =
decoding=E2=80=9D that is required by logical replication. So, we =
mention logical replication in the first place, and to not lose the =
info, mention =E2=80=9Clogical decoding=E2=80=9D in the second place.

>>=20
>> Something like:
>> Note that setting this parameter only enables the required WAL =
format;=20
>> it does not configure logical replication by itself.
>>=20
>=20
> I don't think such note should be added. If the DBA doesn't know how =
to setup
> logical replication, it should read the Logical Replication chapter.
>=20

Agreed.

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/