Re: Allow logical failover slots to wait on synchronous replication

On Sat, Sep 21, 2024 at 6:34 AM John H <johnhyvr(at)gmail(dot)com> wrote:
>
> On Fri, Sep 20, 2024 at 2:44 AM shveta malik <shveta(dot)malik(at)gmail(dot)com> wrote:
> > > >
> > >
> > > The difference is that the purpose of 'synchronized_standby_slots' is
> > > to mention slot names for which the user expects logical walsenders to
> > > wait before sending the logical changes to subscribers. OTOH,
> > > 'synchronous_standby_names' has a different purpose as well. It is not
> > > clear to me if the users would be interested in syncing failover slots
> > > to all the standbys mentioned in 'synchronous_standby_names'.
> > >
> >
> > Okay, I see your point. But not sure what could be the solution here
> > except documenting. But let me think more.
> >
>
> That's a great find. I didn't consider mixed physical and logical
> replicas in synchronous_standby_names.
> I wonder if there are users running synchronous_standby_names with a
> mix of logical and
> physical replicas and what the use case would be.
>

I am also not aware of the actual use cases of mixing physical and
logical synchronous standbys but as we provide that functionality, we
can't ignore it. BTW, I am also not sure if users would like the slots
to be synced on all the standbys mentioned in
synchronous_standby_names. and even, if they are, it is better to have
an explicit way of letting users specify it.

One possible approach is to extend the syntax of
"synchronized_standby_slots" similar to "synchronous_standby_names" so
that users can directly specify slots similarly and avoid waiting for
more than required standbys.

--
With Regards,
Amit Kapila.