Horton

Re: Old row version in hot chain become visible after a freeze

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: "Wood, Dan" <hexpert@amazon.com>
Cc: "pgsql-bugs@postgresql.org" <pgsql-bugs@postgresql.org>
Date: 2017-09-01T00:07:39Z
Lists: pgsql-bugs 
On Thu, Aug 31, 2017 at 4:20 PM, Wood, Dan <hexpert@amazon.com> wrote:
> I was aware of the other one and, in fact, reverted the change just to make sure it wasn’t Involved.  A strange coincidence indeed.

FWIW, the enhanced amcheck does actually detect this:

postgres=# select bt_index_check('t_pkey', true);
ERROR:  XX000: failed to find parent tuple for heap-only tuple at
(0,3) in table "t"
LOCATION:  IndexBuildHeapRangeScan, index.c:2597

The good news is that this happens in a codepath that is also used by REINDEX:

postgres=# reindex index t_pkey ;
ERROR:  XX000: failed to find parent tuple for heap-only tuple at
(0,3) in table "t"
LOCATION:  IndexBuildHeapRangeScan, index.c:2597

Perhaps the lack of field reports of this error suggests that it's
fairly rare. Very hard to be sure about that, though.

-- 
Peter Geoghegan

Commits

  1. Fix freezing of a dead HOT-updated tuple

  2. Fix potential data corruption during freeze

  3. Clarify the contract of partition_rbound_cmp().