Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (1.0)
Subject: Re: Duplicate key value violates unique constraint
 "pg_type_typname_nsp_index" error
From: "Edward J. Sabol" <edwardjsabol@gmail.com>
In-Reply-To: <2325374.1771476717@sss.pgh.pa.us>
Date: Thu, 19 Feb 2026 08:02:33 -0500
Cc: Pgsql-admin <pgsql-admin@lists.postgresql.org>
Message-Id: <ED9777D8-6F59-4B2D-A555-C4574A5B0942@gmail.com>
References: <2325374.1771476717@sss.pgh.pa.us>
To: Tom Lane <tgl@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/ED9777D8-6F59-4B2D-A555-C4574A5B0942%40gmail.com>
Precedence: bulk

On Feb 18, 2026, at 11:51=E2=80=AFPM, Tom Lane wrote:
> =EF=BB=BFEd Sabol <edwardjsabol@gmail.com> writes:
>> Using PostgreSQL 15.16, I have some SQL for creating a materialized view t=
hat's inside a larger transaction which randomly emits the following error:
>> ERROR:  duplicate key value violates unique constraint "pg_type_typname_n=
sp_index"
>> DETAIL:  Key (typname, typnamespace)=3D(largetable, 20488) already exists=
.
>=20
> Huh.  Does it always mention 20488?

Yes, it's always 20488.

>> I've tried the following queries to try to figure out what typnamespace =3D=
 20488 refers to:
>=20
>> SELECT nspname FROM pg_namespace WHERE oid =3D 20488;
>> SELECT count(*) FROM pg_catalog.pg_type WHERE typname =3D 'targetable' AN=
D typnamespace =3D 20488;
>=20
>> Neither query provides any useful information.
>=20
> What do you mean by "any useful information", exactly?  Do these show
> that there is no pg_namespace row with oid =3D 20488?

Well, the first query returns a schema that is completely unrelated to anyth=
ing the SQL is doing. I wondered if the typnamespace =3D 20488 DETAIL was re=
ferring to something else.

> If that is the case, I'd bet on index corruption in the indexes of
> pg_type and/or pg_namespace.  REINDEX might help.

I'll try that. But why would it only happen sometimes?

All of the mentions of this error message I could find on the Internet sugge=
st this happens when you have multiple processes trying to execute the same D=
DL simultaneously, but I'm sure there's only one process executing this tran=
saction.

Thanks,
Ed