Re: pg_stat_activity.waiting_start

On 2017-01-06 10:43:32 -0500, Bruce Momjian wrote:
> On Thu, Jan 5, 2017 at 06:48:17PM -1000, Joel Jacobson wrote:
> > On Thu, Jan 5, 2017 at 4:59 PM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> > > Agreed. No need in adding overhead for short-lived locks because the
> > > milli-second values are going to be meaningless to users. I would be
> > > happy if we could find some weasel value for non-heavyweight locks.
> >
> > To avoid a NULL value for waiting_start, and thanks to non-heavyweight
> > locks don't exceed order-of-milliseconds, I think it would be
> > acceptable to just return now() whenever something wants to know
> > waiting_start i.e. when something selects from pg_stat_activity.
> >
> > The exact value would only be within orders-of-milliseconds away from
> > now() anyway, so one can argue it's not that important, as long as the
> > documentation is clear on that point.
>
> I don't think now() is a good value as it doesn't indicate to the user
> which values are real measurements and which are not. NULL is probably
> the best. +/-infinity is odd too.

Yea. If one wants to make NULL into now() it's trivial enough with a
single coalesce().

Andres