Re: Possibility to disable `ALTER SYSTEM` - Mailing list pgsql-hackers

From Gabriele Bartolini
Subject Re: Possibility to disable `ALTER SYSTEM`
Date
Msg-id CA+VUV5q=za5MqBbXV-OKJc8ukyfQZQjoxSVKvUK4C1wL7gjm2w@mail.gmail.com
Whole thread Raw
In response to Re: Possibility to disable `ALTER SYSTEM`  (Peter Eisentraut <peter@eisentraut.org>)
Responses Re: Possibility to disable `ALTER SYSTEM`
Re: Possibility to disable `ALTER SYSTEM`
List pgsql-hackers
Hi there,

I very much like the idea of a file in the data directory that also controls the copy operations.

Just wanted to highlight though that in our operator we have already applied the read-only postgresql.auto.conf trick to disable the system (see https://cloudnative-pg.io/documentation/current/postgresql_conf/#enabling-alter-system). However, having that file read-only triggered an issue when using pg_rewind to resync a former primary, as pg_rewind immediately bails out when a read-only file is encountered in the PGDATA (see https://github.com/cloudnative-pg/cloudnative-pg/issues/3698).

We might keep this in mind if we go down the path of the separate file.

Thanks,
Gabriele

On Wed, 31 Jan 2024 at 08:43, Peter Eisentraut <peter@eisentraut.org> wrote:
On 31.01.24 06:28, Tom Lane wrote:
>> The idea of adding a file to the data directory appeals to me.
>>
>> optional_runtime_features.conf
>> alter_system=enabled
>> copy_from_program=enabled
>> copy_to_program=disabled
> ... so, exactly what keeps an uncooperative superuser from
> overwriting that file?

The point of this feature would be to keep the honest people honest.

The first thing I did when ALTER SYSTEM came out however many years ago
was to install Nagios checks to warn when postgresql.auto.conf exists.
Because the thing is an attractive nuisance, especially when you want to
do centralized configuration control.  Of course you can bypass it using
COPY PROGRAM etc., but then you *know* that you are *bypassing*
something.  If you just see ALTER SYSTEM, you'll think, "that is
obviously the appropriate tool", and there is no generally accepted way
to communicate that, in particular environment, it might not be.



--
Gabriele Bartolini
Vice President, Cloud Native at EDB

pgsql-hackers by date:

Previous
From: jian he
Date:
Subject: Re: POC, WIP: OR-clause support for indexes
Next
From: Heikki Linnakangas
Date:
Subject: Re: Extending SMgrRelation lifetimes