Horton

Re: BUG #16663: DROP INDEX did not free up disk space: idle connection hold file marked as deleted

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: Pavel Borisov <pashkin.elfe@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2020-12-01T03:06:48Z
Lists: pgsql-bugs, pgsql-hackers 
On Tue, Dec 1, 2020 at 3:55 PM Michael Paquier <michael@paquier.xyz> wrote:
> On Mon, Nov 30, 2020 at 06:59:40PM +1300, Thomas Munro wrote:
> > So I fixed that, by adding a return value to do_truncate() and
> > checking it.  That's the version I plan to commit tomorrow, unless
> > there are further comments or objections.  I've also attached a
> > version suitable for REL_11_STABLE and earlier branches (with a name
> > that cfbot should ignore), where things are slightly different.  In
> > those branches, the register_forget_request() logic is elsewhere.
>
> Hmm.  Sorry for arriving late at the party.  But is that really
> something suitable for a backpatch?  Sure, it is not optimal to not
> truncate all the segments when a transaction dropping a relation
> commits, but this was not completely broken either.

I felt on balance it was a "bug", since it causes operational
difficulties for people and was clearly not our intended behaviour,
and I announced this intention 6 weeks ago.  Of course I'll be happy
to revert it from the back-branches if that's the consensus.  Any
other opinions?

Commits

  1. Use truncate(2) where appropriate.

  2. Free disk space for dropped relations on commit.

  3. Rename the "point is strictly above/below point" comparison operators.