Re: Inconsitancies in pg_stat_bgwriter and pg_stat_database returned values

From: RECHTÉ Marc <marc(dot)rechte(at)meteo(dot)fr>
To: Jehan-Guillaume de Rorthais <jgdr(at)dalibo(dot)com>
Cc: pgsql-bugs(at)lists(dot)postgresql(dot)org
Subject: Re: Inconsitancies in pg_stat_bgwriter and pg_stat_database returned values
Date: 2020-09-17 06:38:10
Message-ID: 746448376.8762238.1600324690698.JavaMail.zimbra@meteo.fr
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

On Wed, 16 Sep 2020 17:06:45 +0200 (CEST)
RECHTÉ Marc <marc(dot)rechte(at)meteo(dot)fr> wrote:

> Hello,
>
> On one particular PG12.3 instance (same behaviour with PG12.4) we are
> experimenting occasional strange values in the above tables.
>
> For instance looping (10s. period) over this request:
>
> SELECT current_timestamp, checkpoints_timed,
> checkpoints_req,
> checkpoint_write_time,
> checkpoint_sync_time,
> buffers_checkpoint,
> buffers_clean,
> maxwritten_clean,
> buffers_backend,
> buffers_backend_fsync,
> buffers_alloc
> FROM pg_stat_bgwriter
>
> Gives: [...]
>
> One can see that some columns that are supposed to increase only are not
> always. For instance checkpoints_timed suddenly jumps from 438 to 3291, then
> back to 438.
>
> We experiment the same issue in pg_stat_database.

----- Mail original -----
De: "Jehan-Guillaume de Rorthais" <jgdr(at)dalibo(dot)com>
À: "RECHTÉ Marc" <marc(dot)rechte(at)meteo(dot)fr>
Cc: pgsql-bugs(at)lists(dot)postgresql(dot)org
Envoyé: Mercredi 16 Septembre 2020 17:31:24
Objet: Re: Inconsitancies in pg_stat_bgwriter and pg_stat_database returned values

The full row are different in your example, not just a few fields. It looks
like mixed values from different instance.

Could you share some more informations about the context and how to reproduce
it?

Regards,

Hello

Thanks a lot, you got it right: there are 2 instances on this machine and they are both configured with:

stats_temp_directory = '/dev/shm'

We some time ago tried, stats_temp_directory = '/dev/shm/instance', but PostgreSQL ignored the subdirectory part and kept creating the temp file at the FS root.
As this is a temporary FS, one cannot in advance create the instance subdirectory as it is wiped out between reboots.

Best regards

In response to

Responses

Browse pgsql-bugs by date

  From Date Subject
Next Message Jehan-Guillaume de Rorthais 2020-09-17 07:39:44 Re: Inconsitancies in pg_stat_bgwriter and pg_stat_database returned values
Previous Message Michael Paquier 2020-09-17 05:53:27 Re: Since '2001-09-09 01:46:40'::timestamp microseconds are lost when extracting epoch