Re: Track the amount of time waiting due to cost_delay

Hi,

On Mon, Dec 16, 2024 at 04:02:56PM -0600, Nathan Bossart wrote:
> On Mon, Dec 16, 2024 at 10:11:23AM +0000, Bertrand Drouvot wrote:
> > +#define PARALLEL_VACUUM_WORKER_DELAY_REPORT_INTERVAL_NS (NS_PER_S)
> >
> > Did not changed in v14, but "PARALLEL_VACUUM_REPORT_INTERVAL_NS" could be
> > an option as well. I think it keeps the key concepts while being more concise (
> > WORKER is somehow implicit in the context).
>
> I think it's important to keep "delay" somewhere in the name, so how about
> PARALLEL_VACUUM_DELAY_REPORT_INTERVAL_NS?

Yeah, sounds good to me (done in the attached).

> > -vacuum_delay_point(void)
> > +static void
> > +vacuum_delay_point_internal(bool is_analyze)
> >
> > Updated the comment on top of it accordingly.
>
> Thanks. I think we need to do some additional adjustments to this
> commentary since external callers should now use
> vacuum/analyze_delay_point().

Agree, I gave it a try in the attached.

Regards,

--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com