On 14 February 2018 at 10:26, Andres Freund <and...@anarazel.de> wrote:

> On 2018-02-14 11:12:30 +0900, Tatsuo Ishii wrote:
> > According to the manual, backend sends a parameter status message when
> > certain configuration variable has been changed and SIGHUP signal is
> sent.
> > https://www.postgresql.org/docs/10/static/protocol-flow.
> html#PROTOCOL-ASYNC
> >
> >   ParameterStatus messages will be generated whenever the active value
> >   changes for any of the parameters the backend believes the frontend
> >   should know about. Most commonly this occurs in response to a SET
> >   SQL command executed by the frontend, and this case is effectively
> >   synchronous ― but it is also possible for parameter status changes
> >   to occur because the administrator changed a configuration file and
> >   then sent the SIGHUP signal to the server.
> >
> > So I connected to PostgreSQL using psql and attached strace to psql.
> > Then I changed standard_conforming_strings and executed pg_ctl
> > reload. The PostgreSQL log shows:
> >
> > 12073 2018-02-14 11:05:22 JST LOG:  received SIGHUP, reloading
> configuration files
> > 12073 2018-02-14 11:05:22 JST LOG:  parameter
> "standard_conforming_strings" changed to "off"
> > 12073 2018-02-14 11:05:22 JST DEBUG:  sending signal 1 to process 12158
> >
> > But as far as strace tells, nothing was sent to psql. Is this expected?
>
> It'll only get sent to the client the next time the server processes a
> query. We can't just at arbitrary points reload the config file or send
> messages out. The SIGHUP handler just sets ConfigReloadPending which
> PostgresMain() then processes:
>
>                 /*
>                  * (6) check for any other interesting events that
> happened while we
>                  * slept.
>                  */
>                 if (ConfigReloadPending)
>                 {
>                         ConfigReloadPending = false;
>                         ProcessConfigFile(PGC_SIGHUP);
>                 }
>
> which'll then, in turn, send out ParameterStatus messages for changed
> GUC_REPORT GUCs.


 I was wondering a while ago - can't we just set our own proc's latch here,
so we wake up and send it earlier if we're in the idle main loop?

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

Reply via email to