Re: [bug fix] ALTER TABLE SET LOGGED/UNLOGGED on a partitioned table does nothing silently

On 2020-Dec-08, tsunakawa(dot)takay(at)fujitsu(dot)com wrote:

> From: Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org>
> > Does "ALTER TABLE ONLY parent" work correctly? Namely, do not affect
> > existing partitions, but cause future partitions to acquire the new
> > setting.
>
> Yes, it works correctly in the sense that ALTER TABLE ONLY on a
> partitioned table does nothing because it has no storage and therefore
> logged/unlogged has no meaning.

But what happens when you create another partition after you change the
"loggedness" of the partitioned table?

> > This sounds very much related to previous discussion on REPLICA IDENTITY
> > not propagating to partitions; see
> > https://postgr.es/m/201902041630.gpadougzab7v@alvherre.pgsql
> > particularly Robert Haas' comments at
> > http://postgr.es/m/CA+TgmoYjksObOzY8b1U1=BsP_m+702eOf42fqRtQTYio
> > 1NunbA(at)mail(dot)gmail(dot)com
> > and downstream discussion from there.
>
> There was a hot discussion. I've read through it. I hope I'm not
> doing strange in light of that discussioin.

Well, my main take from that is we should strive to change all
subcommands together, in some principled way, rather than change only
one or some, in arbitrary ways.