Date: Tue, 26 Mar 2024 14:09:11 +0100 (CET)
From: gparc@free.fr
To: pgsql-bugs <pgsql-bugs@lists.postgresql.org>
Message-ID: <621556601.826991929.1711458551936.JavaMail.zimbra@free.fr>
In-Reply-To: <18402-23452305346476d6@postgresql.org>
References: <18402-23452305346476d6@postgresql.org>
Subject: Re: BUG #18402: Attaching a new partition doesn't reuse the
 prebuilt index on said partition
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Thread-Topic: BUG #18402: Attaching a new partition doesn't reuse the prebuilt
 index on said partition
Thread-Index: MoR/qSVpV+/BNAUa9xjw0seRFYxfrQ==
Archived-At: 
 <https://www.postgresql.org/message-id/621556601.826991929.1711458551936.JavaMail.zimbra%40free.fr>
Precedence: bulk


Hello,

is there a week retention before this report is delivered to the mailing li=
st ?

Regards
Gilles

----- Mail original -----
> De: "PG Bug reporting form" <noreply@postgresql.org>
> =C3=80: "pgsql-bugs" <pgsql-bugs@lists.postgresql.org>
> Cc: "gparc" <gparc@online.fr>
> Envoy=C3=A9: Mercredi 20 Mars 2024 15:29:14
> Objet: BUG #18402: Attaching a new partition doesn't reuse the prebuilt i=
ndex on said partition

> The following bug has been logged on the website:
>=20
> Bug reference:      18402
> Logged by:          Gilles PARC
> Email address:      gparc@online.fr
> PostgreSQL version: 16.2
> Operating system:   Linux
> Description:
>=20
> Example :
>=20
> First I create a list partitioned table and fill it
> CREATE TABLE master
> (
> id integer NOT NULL,
> millesime varchar(4) NOT NULL,
> CONSTRAINT master_pk PRIMARY KEY (id, millesime)
> ) partition by list (millesime);
>=20
> CREATE TABLE master_2022 PARTITION OF master FOR VALUES IN ('2022');
> CREATE TABLE master_2023 PARTITION OF master FOR VALUES IN ('2023');
> CREATE TABLE master_2024 PARTITION OF master FOR VALUES IN ('2024');
> CREATE TABLE master_2025 PARTITION OF master FOR VALUES IN ('2025');
>=20
> DO
> $do$
> BEGIN
>   FOR i IN 2022..2025 LOOP
>      INSERT INTO master VALUES (generate_series(1,100000), i::varchar);
>   END LOOP;
> END
> $do$;
>=20
> I now list the created indexes:
> \diS+ master*
>                                                      List of relations
> Schema |       Name       |       Type        |  Owner   |    Table    |
> Persistence | Access method |  Size   | Description
> --------+------------------+-------------------+----------+-------------+=
-------------+---------------+---------+-------------
> public | master_2022_pkey | index             | postgres | master_2022 |
> permanent   | btree         | 3088 kB |
> public | master_2023_pkey | index             | postgres | master_2023 |
> permanent   | btree         | 3088 kB |
> public | master_2024_pkey | index             | postgres | master_2024 |
> permanent   | btree         | 3088 kB |
> public | master_2025_pkey | index             | postgres | master_2025 |
> permanent   | btree         | 3088 kB |
> public | master_pk        | partitioned index | postgres | master      |
> permanent   | btree         | 0 bytes |
> (5 rows)
>=20
>=20
> Now I prepare a new 2026 partition with correct constraints and indexes a=
nd
> attach it :
>=20
> CREATE TABLE master_2026 (LIKE master INCLUDING DEFAULTS INCLUDING
> CONSTRAINTS);
> ALTER TABLE master_2026 ADD CONSTRAINT master_2026_ck CHECK (millesime =
=3D
> '2026');
> DO
> $do$
> BEGIN
>      INSERT INTO master_2026 VALUES (generate_series(1,100000), '2026');
> END
> $do$;
> CREATE UNIQUE INDEX CONCURRENTLY master_2026_pkey ON master_2026 (id,
> millesime);
> ALTER TABLE master ATTACH PARTITION master_2026 FOR VALUES IN ('2026');
> ALTER TABLE master_2026 DROP CONSTRAINT master_2026_ck;
>=20
> But now, when I list the indexes again, I see that my unique index above =
is
> not considered to use
> and a new one is created (master_2026_pkey1).
> \diS+ master*
> Schema |       Name        |       Type        |  Owner   |    Table    |
> Persistence | Access method |  Size   | Description
> --------+-------------------+-------------------+----------+-------------=
+-------------+---------------+---------+-------------
> public | master_2022_pkey  | index             | postgres | master_2022 |
> permanent   | btree         | 3088 kB |
> public | master_2023_pkey  | index             | postgres | master_2023 |
> permanent   | btree         | 3088 kB |
> public | master_2024_pkey  | index             | postgres | master_2024 |
> permanent   | btree         | 3088 kB |
> public | master_2025_pkey  | index             | postgres | master_2025 |
> permanent   | btree         | 3088 kB |
> public | master_2026_pkey  | index             | postgres | master_2026 |
> permanent   | btree         | 3104 kB |
> public | master_2026_pkey1 | index             | postgres | master_2026 |
> permanent   | btree         | 3104 kB |
> public | master_pk         | partitioned index | postgres | master      |
> permanent   | btree         | 0 bytes |
> (7 rows)
>=20
> why is it so ?
>=20
> Bug or Feature ?
> Regards
> Gilles