Re: Reloading pgbouncer and systemd

Le Wed, 30 Mar 2016 16:15:00 +0300,
Devrim Gündüz <devrim(at)gunduz(dot)org> a écrit :

> Looks good to me. Can you please create a ticket for this, so that we can keep
> track?

Done: https://redmine.postgresql.org/issues/1042

Thank you Devrim!

> On Mon, 2016-02-22 at 17:38 +0100, Jehan-Guillaume de Rorthais wrote:
> > Hello,
> >
> > While working with pgbouncer under CentOS7, I noticed an incompatibility
> > between pgbouncer and systemd when reloading.
> >
> > The current reload command is:
> >
> >   ExecReload=/usr/bin/pgbouncer -R -q ${BOUNCERCONF}
> >
> > But this command create a new pgbouncer process...and keeps it in
> > foreground. As the service type is set as "forking", in this situation
> > systemctl just sits
> > and wait for a return code (maybe there's a timeout, but I did not have the
> > patience to wait for it).
> >
> > Either add a '-d' to this command, or send a HUP signal to the process.
> >
> > I prefer the second option as it doesn't create a new process that takes
> > over the sockets, but just reloads the conf file as expected. Much lighter.
> >
> > So, imho, the pgbouncer.service file should have:
> >
> >   ExecReload=/bin/kill -HUP $MAINPID