From: Tom Lane <tgl@sss.pgh.pa.us>
To: mathiaskunter@gmail.com
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #17347: pg_upgrade: analyze_new_cluster script analyzes wrong
 cluster
In-reply-to: <17347-27b988f51d9dee1a@postgresql.org>
References: <17347-27b988f51d9dee1a@postgresql.org>
Comments: In-reply-to PG Bug reporting form <noreply@postgresql.org>
	message dated "Tue, 28 Dec 2021 09:47:42 +0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1760673.1640707235.1@sss.pgh.pa.us>
Date: Tue, 28 Dec 2021 11:00:35 -0500
Message-ID: <1760674.1640707235@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1760674.1640707235%40sss.pgh.pa.us>
Precedence: bulk

PG Bug reporting form <noreply@postgresql.org> writes:
> When upgrading a PostgreSQL cluster with pg_upgrade (tested while upgrading
> from 11.14 to 12.9, but supposedly affects all versions), and the --new-port
> argument is used to specify a new port number for the upgraded cluster, then
> the generated analyze_new_cluster script will connect to and analyze a
> different cluster instead of the upgraded one.
> The vacuumdb command found within the analyze_new_cluster script should
> explicitly specify the port number of the upgraded cluster, so that it will
> actually analyze the correct cluster after the upgrade.

Hmm.  I'm not sure whether changing that would be an improvement.
The assumption you're making is that --new-port will be the new
cluster's port number after it's started for real.  But it seems
at least as likely that it's just a nonce specification used for
long enough to finish sucking data out of your old cluster, and
then you'll start the new server in place of your old one (ie
using the old port).

It's a bit moot now anyway, since pg_upgrade doesn't create
that script anymore in v14 and up.

			regards, tom lane