Re: Read-only connection mode for AI workflows.
Peter Eisentraut <peter@eisentraut.org>
From: Peter Eisentraut <peter@eisentraut.org>
To: Andrei Lepikhov <lepihov@gmail.com>, Jack Bonatakis <jack@bonatak.is>,
pgsql-hackers <pgsql-hackers@postgresql.org>,
Bruce Momjian <bruce.momjian@enterprisedb.com>,
Andres Freund <andres@anarazel.de>
Date: 2026-03-19T10:09:26Z
Lists: pgsql-hackers
On 19.03.26 08:44, Andrei Lepikhov wrote: > On 16/3/26 22:01, Andrei Lepikhov wrote: >> On 16/3/26 20:28, Jack Bonatakis wrote: >>> On Mon, Mar 16, 2026, at 2:08 PM, Andrei Lepikhov wrote: >>>> I believe the pg_readonly [1] extension does what you're looking >>>> for, so >>>> you might want to give it a try. >>> Please correct me if I am mistaken, but it looks like pg_readonly >>> operates at the database or cluster level. > > Take a look at the [1] project. It's a simpler version of [2] that > always switches to read-only mode. > To use it, just have your connection pooler load the 'safesession' > module. This will keep the session in read-only mode until it ends. > There are no GUCs, and there is no way to change the mode, even for a > superuser. Does this seem safe enough? > > We could improve it by restricting manual calls to specific utility > operations, such as VACUUM or REINDEX. However, we would need some > specifications first. Here is a stalled project to implement ALTER SYSTEM READ ONLY: https://www.postgresql.org/message-id/flat/CAAJ_b97KZzdJsffwRK7w0XU5HnXkcgKgTR69t8cOZztsyXjkQw%40mail.gmail.com Maybe that could be reactivated, but the reason it stalled was that it was hard.