Horton

Re: [BUG] non archived WAL removed during production crash recovery

Kyotaro Horiguchi <horikyota.ntt@gmail.com>

View thread View on postgr.es
From: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
To: michael@paquier.xyz
Cc: jgdr@dalibo.com, masao.fujii@oss.nttdata.com, pgsql-bugs@lists.postgresql.org
Date: 2020-04-23T05:05:46Z
Lists: pgsql-bugs, pgsql-hackers 
FWIW, the test for 10- looks fine, but I have one qustion.

+ 	'archive success reported in pg_stat_archiver for WAL segment $segment_name_

This seems intending to show an actual segment name in the message,
but it is really shown as "... WAL segment $segment_name_1". Is that
intended?

At Thu, 23 Apr 2020 08:46:18 +0900, Michael Paquier <michael@paquier.xyz> wrote in 
> On Wed, Apr 22, 2020 at 06:17:17PM +0200, Jehan-Guillaume de Rorthais wrote:
> > I found an extra useless line of code in v9 patch. Please, find in
> > attachment v10. Sorry for this.
> 
> Thanks for helping here, your changes make sense.  This looks mostly
> fine to me except that part:
> +$standby1->poll_query_until('postgres',
> +   qq{ SELECT pg_xlog_location_diff('$primary_lsn', pg_last_xlog_replay_location()) = 0 })
> +  or die "Timed out while waiting for xlog replay";
> Here we should check if $primary_lsn is at least
> pg_last_xlog_replay_location().  Checking for an equality may stuck
> the test if more WAL gets replayed.  For example you could have a
> concurrent autovacuum generating WAL.

Autovacuum is turned off in this case, but anyway other kinds of WAL
records can be generated.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center

Commits

  1. Fix interpolation in test name.

  2. Add more TAP coverage for archive status with crash recovery of standbys

  3. Fix handling of WAL segments ready to be archived during crash recovery

  4. Fix WAL recycling on standbys depending on archive_mode