| From: | Justin Pryzby <pryzby(at)telsasoft(dot)com> |
|---|---|
| To: | Christoph Berg <myon(at)debian(dot)org>, Devrim Gündüz <devrim(at)gunduz(dot)org>, pgsql-pkg-yum(at)postgresql(dot)org |
| Subject: | Re: [PostgreSQL YUM Repository - Bug #6855] please configure systemd to not sigkill the postmaster |
| Date: | 2021-11-30 16:13:20 |
| Message-ID: | 20211130161320.GT17618@telsasoft.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-pkg-yum |
On Tue, Nov 30, 2021 at 09:07:31AM +0100, Christoph Berg wrote:
> Re: Justin Pryzby
> > > If it stops slowly and hits the stop timeout, systemd sends sigkill, which then makes the restart extra special painful and slow.
> > >
> > > The debs have: TimeoutStopSec=1h
> > >
> > > Use TimeoutStopSec=infinity or SendSIGKILL=no ?
>
> Fwiw, I didn't use infinity for stopping as I was afraid that might
> mean services in a weird state might prevent servers from rebooting
> indefinitely. Better have them recover after one hour than having to
> drive to the DC.
Thanks.
I think either way could be reasonable.
The initscripts are customizable, anyway.
If something is wrong and storage performance is terrible, once the DB starts
shutting down, sending SIGKILL could make everything worse, since then the
startup process will have to fsync() every file and replay WAL.
--
Justin
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Hasan Marzooq | 2021-12-06 14:38:34 | Installing plv89 extension to Postgres 13 |
| Previous Message | Christoph Berg | 2021-11-30 08:07:31 | Re: [PostgreSQL YUM Repository - Bug #6855] please configure systemd to not sigkill the postmaster |