Date: Mon, 22 Apr 2019 14:16:28 -0400
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Robert Haas <robertmhaas@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>,
	David Rowley <david.rowley@2ndquadrant.com>,
	Andres Freund <andres@anarazel.de>,
	PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: pg_dump is broken for partition tablespaces
Message-ID: <20190422181628.GA29329@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: 
 <CA+Tgmoa9NsQxwKcqpq-UwYfJ3nYKWFs+p+GYBhq-F_phmUv_VA@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On 2019-Apr-22, Robert Haas wrote:

> PostgreSQL has historically and very deliberately *not made a
> distinction* between "this object is in the default tablespace" and
> "this object is in tablespace X which happens to be the default."  I
> think that it's too late to invent such a distinction for reasons of
> backward compatibility -- and if we were going to do it, surely it
> would need to exist for both partitioned tables and the partitions
> themselves. Otherwise it just produces more strange inconsistencies.

Yeah, this is probably right.  (I don't think it's the same thing that
Tom was saying, though, or at least I didn't understand his argument
this way.)

I think we can get out of this whole class of problems by forbidding the
TABLESPACE clause for partitioned rels from mentioning the database
tablespace -- that is, users either mention some *other* tablespace, or
partitions follow default_tablespace like everybody else.  AFAICS with
that restriction this whole problem does not arise, and the patch may
become simpler.  I'll give it a spin.

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