Re: Pgbouncer discard all - Mailing list pgsql-general

From Nicola Contu
Subject Re: Pgbouncer discard all
Date
Msg-id CAMTZZh1g22EbCqofHZq4fsDaP+yFkWmsnnLai7mY=dUTH_n+kQ@mail.gmail.com
Whole thread Raw
In response to Re: Pgbouncer discard all  (Martín Marqués <martin.marques@2ndquadrant.com>)
Responses Re: Pgbouncer discard all
RE: Pgbouncer discard all
List pgsql-general
 2492534808 | dev    |   7355 | 1833427130 | pgbouncer       |                         | 10.151.2.145   |                 |       60570 | 2018-10-16 14:13:05.151015+00 |                               | 2018-10-16 15:10:40.309993+00 | 2018-10-16 15:10:40.310038+00 | Client
  | ClientRead          | idle   |             |              | DISCARD ALL

They are idle actually.
Will they count as client connection on the total amount set on the postgres.conf?

Il giorno mar 16 ott 2018 alle ore 16:22 Martín Marqués <martin.marques@2ndquadrant.com> ha scritto:
El 16/10/18 a las 09:59, Scot Kreienkamp escribió:
> Are you sure they’re actually waiting?  Don’t forget 10.5 will show the
> last query executed even if the connection is idle.  I believe discard
> all would be the last command the pgbouncer would send to the database
> when the client is done as it resets the connection for the next
> client.  So what you’re describing would seem to be expected behavior. 

He might have been referring to client waiting. That is visible in the
pgbouncer pseudo-database

OTOH if the the waiting is seen in pg_stat_activity, then pgbouncer has
nothing to do. The connection has already been assigned to the client
and the waiting is happening on the database server, not the pooler.

Regards,

--
Martín Marqués                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services

pgsql-general by date:

Previous
From: Nicola Contu
Date:
Subject: Re: Swap on postgres master server
Next
From: Martin Marques
Date:
Subject: Re: Pgbouncer discard all