Re: partitioned indexes and tablespaces

On 2018-Nov-02, Amit Langote wrote:

> On 2018/11/02 10:27, Michael Paquier wrote:

> > It seems to me that the current behavior is wanted in this case, because
> > partitioned tables and partitioned indexes have no physical storage.
>
> Keith Fiske complained about this behavior for partitioned *tables* a few
> months ago, which led to the following discussion:
>
> https://www.postgresql.org/message-id/flat/CAKJS1f9PXYcT%2Bj%3DoyL-Lquz%3DScNwpRtmD7u9svLASUygBdbN8w%40mail.gmail.com
>
> It's Michael's message that was the last one on that thread. :)
>
> https://www.postgresql.org/message-id/20180413224007.GB27295%40paquier.xyz

I agree with Fiske, FWIW. I think the current behavior results because
people (including me) overlooked things, not because it was designed
explicitly that way.

> By the way, if we decide to do something about this, I think we do the
> same for partitioned tables.

I'm up for changing the behavior of partitioned tables in pg12 (please
send a patch), but I'm up for changing the behavior of partitioned
tables in pg11.

> There are more than one interesting
> behaviors possible that are mentioned in the above thread for when
> parent's reltablespace is set/changed.

I'm *NOT* proposing to move existing partitions to another tablespace,
in any case.

> IOW, I agree with Michael that if something will be back-patched to 11, it
> should be a small patch to make the unsupported relkind error go away.

I don't.

--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services