Date: Tue, 23 Mar 2021 12:25:23 -0300
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Amit Langote <amitlangote09@gmail.com>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: ALTER TABLE .. DETACH PARTITION CONCURRENTLY
Message-ID: <20210323152523.GA9721@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20210323145554.GA7700@alvherre.pgsql>
User-Agent: Mutt/1.10.1 (2018-07-13)
Archived-At: 
 <https://www.postgresql.org/message-id/20210323152523.GA9721%40alvherre.pgsql>
Precedence: bulk

I'm coming around to the idea that the fact that you can cancel the wait
phase of DETACH CONCURRENTLY creates quite a disaster, and it's not easy
to get away from it.  The idea that REPEATABLE READ mode means that you
now see detached partitions as if they were in normal condition, is
completely at odds with that behavior. 

I think a possible solution to this problem is that the "detach" flag in
pg_inherits is not a boolean anymore, but an Xid (or maybe two Xids).
Not sure exactly which Xid(s) yet, and I'm not sure what are the exact
rules, but the Xid becomes a marker that indicates an horizon past which
the partition is no longer visible.  Then, REPEATABLE READ can see the
partition, but only if its snapshot is older than the Xid.

-- 
Álvaro Herrera       Valdivia, Chile
"La persona que no quería pecar / estaba obligada a sentarse
 en duras y empinadas sillas    / desprovistas, por cierto
 de blandos atenuantes"                          (Patricio Vogel)