From: Antonin Houska <ah@cybertec.at>
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
cc: Marcos Pegoraro <marcos@f10.com.br>, Michael Banck <mbanck@gmx.net>,
    Junwang Zhao <zhjwpku@gmail.com>,
    Kirill Reshke <reshkekirill@gmail.com>,
    Pavel Stehule <pavel.stehule@gmail.com>,
    Michael Paquier <michael@paquier.xyz>,
    PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Subject: Re: why there is not VACUUM FULL CONCURRENTLY?
In-reply-to: <202503031807.dnacvpgnjkz7@alvherre.pgsql>
References: <202503031807.dnacvpgnjkz7@alvherre.pgsql>
Comments: In-reply-to Alvaro Herrera <alvherre@alvh.no-ip.org>
   message dated "Mon, 03 Mar 2025 19:07:00 +0100."
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <49791.1741027227.1@localhost>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 03 Mar 2025 19:40:27 +0100
Message-ID: <49792.1741027227@localhost>
Archived-At: 
 <https://www.postgresql.org/message-id/49792.1741027227%40localhost>
Precedence: bulk

Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:

> On 2025-Feb-26, Antonin Houska wrote:
> =

> > @@ -403,39 +381,38 @@ cluster_rel(Relation OldHeap, Oid indexOid, Clus=
terParams *params)
> >  	 * would work in most respects, but the index would only get marked =
as
> >  	 * indisclustered in the current database, leading to unexpected beh=
avior
> >  	 * if CLUSTER were later invoked in another database.
> > +	 *
> > +	 * REPACK does not set indisclustered. XXX Not sure I understand the
> > +	 * comment above: how can an attribute be set "only in the current
> > +	 * database"?
> >  	 */
> =

> Regarding this XXX comment, what's going on here is this: a CLUSTER
> command needs to remember the index that a table is clustered on.  We
> keep track of this in pg_index.indisclustered.  But pg_index is a local
> relation, not shared across databases -- so the current CLUSTER command
> can effect the update on the current database's pg_index only, not on
> other databases.  So if the user were to run CLUSTER on one database
> specifying an index, then connect to another one and expect CLUSTER
> without specifying an index to honor the previously specified index,
> that would not work.  Naturally this is only a problem for shared
> catalogs.  Not being able to handle this for shared catalogs is not a
> big loss.

Thanks for explanation. The reason I failed to understand this was probabl=
y
that I tried to imagine something worse.

-- =

Antonin Houska
Web: https://www.cybertec-postgresql.com