MIME-Version: 1.0
References: <CALDaNm2Dxmhq08nr4P6G+24QvdBo_GAVyZ_Q1TcGYK+8NHs9xw@mail.gmail.com>
 <CAA4eK1+Dw+0B9vb6rTcPmFWJHyksfu8LJwph1_01f8+C63+DnQ@mail.gmail.com>
In-Reply-To: <CAA4eK1+Dw+0B9vb6rTcPmFWJHyksfu8LJwph1_01f8+C63+DnQ@mail.gmail.com>
From: vignesh C <vignesh21@gmail.com>
Date: Mon, 25 Sep 2023 00:32:03 +0530
Message-ID: <CALDaNm2JDxrsSnX21rnVZNAmZ7xcVE5vb2vrMMnPsU1Xrhz3BA@mail.gmail.com>
Subject: Re: Invalidate the subscription worker in cases where a user loses
 their superuser status
To: Amit Kapila <amit.kapila16@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: <https://www.postgresql.org/message-id/CALDaNm2JDxrsSnX21rnVZNAmZ7xcVE5vb2vrMMnPsU1Xrhz3BA%40mail.gmail.com>
Precedence: bulk

On Sat, 23 Sept 2023 at 11:28, Amit Kapila <amit.kapila16@gmail.com> wrote:
>
> On Sat, Sep 23, 2023 at 1:27=E2=80=AFAM vignesh C <vignesh21@gmail.com> w=
rote:
> >
> >
> > Fixed this issue by checking if the subscription owner has changed
> > from superuser to non-superuser in case the pg_authid rows changes.
> > The attached patch has the changes for the same.
> >
>
> @@ -3952,7 +3953,9 @@ maybe_reread_subscription(void)
>   newsub->passwordrequired !=3D MySubscription->passwordrequired ||
>   strcmp(newsub->origin, MySubscription->origin) !=3D 0 ||
>   newsub->owner !=3D MySubscription->owner ||
> - !equal(newsub->publications, MySubscription->publications))
> + !equal(newsub->publications, MySubscription->publications) ||
> + (!superuser_arg(MySubscription->owner) &&
> + MySubscription->isownersuperuser))
>   {
>   if (am_parallel_apply_worker())
>   ereport(LOG,
> @@ -4605,6 +4608,13 @@ InitializeLogRepWorker(void)
>   proc_exit(0);
>   }
>
> + /*
> + * Fetch subscription owner is a superuser. This value will be later
> + * checked to see when there is any change with this role and the worker
> + * will be restarted if required.
> + */
> + MySubscription->isownersuperuser =3D superuser_arg(MySubscription->owne=
r);
>
> Why didn't you filled this parameter in GetSubscription() like other
> parameters? If we do that then the comparison of first change in your
> patch will look similar to all other comparisons.

I felt this variable need not be added to the pg_subscription catalog
table, instead we could save the state of subscription owner when the
worker is started and compare this value during invalidations. As this
information is added only to the memory Subscription structure and not
added to the catalog FormData_pg_subscription, the checking is
slightly different in this case. Also since this variable will be used
only within the worker, I felt we need not add it to the catalog.

Regards,
Vignesh