On Mon, 15 Apr 2019 at 05:32, Alvaro Herrera <alvherre@2ndquadrant.com> wrote:
>
> On 2019-Apr-14, Andres Freund wrote:
>
> > On 2019-04-14 10:38:05 -0400, Tom Lane wrote:
> > > It's entirely possible BTW that this whole business of inheriting
> > > tablespace from the partitioned table is broken and should be thrown
> > > out. I certainly don't see any compelling reason for partitions to
> > > act differently from regular tables in this respect, and the more
> > > problems we find with the idea, the less attractive it seems.
> >
> > Indeed. After discovering during the tableam work, and trying to write
> > tests for the equivalent feature for tableam, I decided that just not
> > allowing AM specifications for partitioned tables is the right call - at
> > least until the desired behaviour is clearer. The discussion of the last
> > few days makes me think so even more.
>
> To be honest, when doing that feature I neglected to pay attention to
> (read: forgot about) default_tablespace, and it's mostly the
> interactions with that feature that makes this partitioned table stuff
> so complicated. I'm not 100% convinced yet that we need to throw it out
> completely, but I'm less sure now about it than I was before.
FWIW, I was trying to hint in [1] that this all might be more trouble
than its worth.
To be honest, if I'd done a better job of thinking through the
implications of this tablespace inheritance in ca4103025d, then I'd
probably have not bothered submitting a patch for it. We could easily
revert that, but we'd still be left with the same behaviour in
partitioned indexes, which is in PG11.
[1] https://www.postgresql.org/message-id/CAKJS1f-52x3o16fsd4=tBPKct9_E0uEg0LmzOgxBqLiuZsj-SA@mail.gmail.com
--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services