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: "denni.pat" <denni.pat@gmail.com>
Cc: pgsql-bugs@postgresql.org
Date: 2020-10-15T23:54:32Z
Lists: pgsql-bugs, pgsql-hackers 
On Thu, Oct 15, 2020 at 8:20 PM denni.pat <denni.pat@gmail.com> wrote:
> thanks for the patch.
> Do you think it can be included in the next minor releases or the only
> solution will be to recompile?

I would vote +1 for back-patching a fix for this problem (that is,
pushing it into the minor releases), because I agree that it's very
arguably a bug that we treat the segments differently, and looking
around I do see reports of people having to terminate processes to get
their disk space back.  I'd definitely want a consensus on that plan
from some experienced reviewers and testers, though.  For anyone
wanting to test this, you might want to set RELSEGSIZE to a smaller
number in src/include/pg_config.h.

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.