Re: Adding a '--clean-publisher-objects' option to 'pg_createsubscriber' utility. - Mailing list pgsql-hackers

On Wed, 5 Feb 2025 at 07:49, Hayato Kuroda (Fujitsu)
<kuroda.hayato@fujitsu.com> wrote:
>
> Dear Shlok,
>
> > 4. Instead of warning we should throw an error here:
> > +       if (PQresultStatus(res) != PGRES_TUPLES_OK)
> > +       {
> > +           pg_log_warning("could not obtain publication information: %s",
> > +                          PQresultErrorMessage(res));
> > +
>
> I don't think so. ERROR evokes user to retry the command or recreate the physical
> replica, but the conversion has already been finished when drop_all_publications()
> is called. Cleanup operations should not affect the final result.
> drop_primary_replication_slot() and drop_failover_replication_slots() raise WARNING
> when they fail to drop objects because they are just cleanup functions.
> I feel we can follow this manner.
>

Hi Kuroda-san,

I agree with you. Raising WARNING makes sense to me.

Thanks and Regards,
Shlok Kyal