Horton

Re: [HACKERS] ginInsertCleanup called from vacuum could still miss tuples to be deleted

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: Robert Haas <robertmhaas@gmail.com>
Cc: Masahiko Sawada <sawada.mshk@gmail.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2017-11-16T20:34:50Z
Lists: pgsql-hackers 
On Thu, Nov 16, 2017 at 12:29 PM, Robert Haas <robertmhaas@gmail.com> wrote:
> I am also clear on what the consequences of this bug are.  It seems
> like it could harm insert performance by making us wait when we
> shouldn't, but can it cause corruption?  That I'm not sure about.  If
> there's already a cleanup of the pending list in progress, how do
> things go wrong?

While it was probably the right thing for you to go ahead and commit
the patch, I think that the bug undermines whatever confidence there
was in the correctness of the GIN pending list cleanup code. I have
far more questions than answers right now, though.


-- 
Peter Geoghegan

Commits

  1. Fix broken cleanup interlock for GIN pending list.

  2. Prevent multiple cleanup process for pending list in GIN.

  3. Add pages deleted from pending list to FSM

  4. Fix race condition in GIN posting tree page deletion.