Horton

Re: [PATCH] Btree BackwardScan race condition on Standby during VACUUM

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: "Andrey M. Borodin" <x4mmm@yandex-team.ru>
Cc: Michail Nikolaev <michail.nikolaev@gmail.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2020-03-17T19:37:57Z
Lists: pgsql-hackers 
On Mon, Mar 16, 2020 at 10:20 PM Andrey M. Borodin <x4mmm@yandex-team.ru> wrote:
> It seems to me that it's exactly the same check that I was trying to verify in amcheck patch [0].
> But there it was verified inside amcheck, but here it is verified by index scan.

Maybe we can accept your patch after fixing this bug. My objection to
the patch was that it couples locks in a way that's not compatible
with btree_xlog_unlink_page(). But the problem now seems to have been
btree_xlog_unlink_page() itself. It's possible that there are problems
elsewhere, but my recollection is that btree_xlog_unlink_page() was
the problem.

-- 
Peter Geoghegan

Commits

  1. Fix replica backward scan race condition.

  2. Justify nbtree page split locking in code comment.

  3. Remove unneeded "pin scan" nbtree VACUUM code.

  4. Compute XID horizon for page level index vacuum on primary.

  5. Make heap TID a tiebreaker nbtree index column.