Re: Support to define custom wait events for extensions

Masahiro Ikeda <ikedamsh@oss.nttdata.com>

From: Masahiro Ikeda <ikedamsh@oss.nttdata.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: pgsql-hackers@lists.postgresql.org
Date: 2023-06-16T11:44:53Z
Lists: pgsql-hackers
On 2023-06-16 16:46, Michael Paquier wrote:
> On Fri, Jun 16, 2023 at 11:14:05AM +0900, Masahiro Ikeda wrote:
>> I tried to query on pg_stat_activity to check the background worker
>> invoked by pg_prewarm. But, I found that pg_stat_activity doesn't show
>> it although I may be missing something...
>> 
>> So, I tried to implement TAP tests. But I have a problem with it.
>> I couldn't find the way to check the status of another backend
>> while the another backend wait with custom wait events.
> 
> Hmm.  Right.  It seems to me that BGWORKER_BACKEND_DATABASE_CONNECTION
> is required in this case, with BackgroundWorkerInitializeConnection()
> to connect to a database (or not, like the logical replication
> launcher if only access to shared catalogs is wanted).
> 
> I have missed that the leader process of pg_prewarm does not use that,
> because it has no need to connect to a database, but its workers do.
> So it is not going to show up in pg_stat_activity.

Yes. Thanks to your advice, I understood that
BGWORKER_BACKEND_DATABASE_CONNECTION is the reason.

I could make the TAP test that invokes a background worker waiting 
forever
and checks its custom wait event in pg_stat_activity. So, I'll make 
patches
including test codes next week.

Regards,
-- 
Masahiro Ikeda
NTT DATA CORPORATION



Commits

  1. Change custom wait events to use dynamic shared hash tables

  2. Support custom wait events for wait event type "Extension"

  3. Add WAIT_EVENT_{CLASS,ID}_MASK in wait_event.c

  4. worker_spi: Switch to TAP tests

  5. Simplify some conditions related to [LW]Lock in generate-wait_event_types.pl