Re: Add GoAway protocol message for graceful but fast server shutdown/switchover

From: Jelte Fennema-Nio <postgres(at)jeltef(dot)nl>
To: Tatsuo Ishii <ishii(at)postgresql(dot)org>
Cc: jacob(dot)champion(at)enterprisedb(dot)com, tomas(at)vondra(dot)me, zsolt(dot)parragi(at)percona(dot)com, pgsql-hackers(at)lists(dot)postgresql(dot)org, davecramer(at)gmail(dot)com, hlinnaka(at)iki(dot)fi
Subject: Re: Add GoAway protocol message for graceful but fast server shutdown/switchover
Date: 2026-03-31 07:12:09
Message-ID: CAGECzQSbr=WDJDZ-fFTDRjY7udC6H9Sb3VHWk9BxWx9uTG27tg@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Tue, 31 Mar 2026 at 01:17, Tatsuo Ishii <ishii(at)postgresql(dot)org> wrote:
> Interesting. In Pgpool-II case, client disconnection is not a problem,
> the connection to PostgreSQL is kept open anyway (if connection
> expiration is not set). Probably this is because Pgpool-II only
> supports "session level" connection pooling.

I guess, I was not clear. Because the same problem exists for
Pgpool-II. When pgpool forwards the message to the client, the client
will disconnect from pgpool. But if connection_life_time is not
reached (and by default this is unlimited) pgpool will not disconnect
from the postgres server. So the postgres server has not actually
achieved the intended goal, but the client still disconnected for no
benefit, only downsides.

Maybe the problem is not too bad, i.e. it will cause some unnecessary
disconnects from the client, but it shouldn't cause big problems. So
maybe the ParameterStatus approach is worth exploring again (I
remember I ran into some problems, due to us normally only sending
ParameterStatus at the end of a query, but I'm pretty sure that can be
worked around somehow).

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Jelte Fennema-Nio 2026-03-31 07:16:40 Re: Improve pgindent's formatting named fields in struct literals and varidic functions
Previous Message Bertrand Drouvot 2026-03-31 07:10:51 Re: Adding locks statistics