[Update] Explain -infinity of next_start in jobs. about docs

[NunoFilipeSantos]

This is the only hit I read on docs: The next time at which to run the job. The job can be paused by setting this value to infinity, and restarted with a value of now().

@mkindahl or @erimatnor might know more.

[mkindahl]

@Loquacity When the job starts execution, it seems to set the next_start value to -infinity. Once the job completes, it will set the value to the actual computed next start time. Look in bgw_job_stat_tuple_mark_start and you will se that it is set to DT_NOBEGIN, which is a constant representing -infinity.

[mindrunner]

Me too wondering what that means.

[mindrunner]

I am guessing here, but it seems that it means something like:

Never ran, to be scheduled as soon as possible

[Loquacity]

@NunoFilipeSantos Can you help here?

[mindrunner]

So if a running Job gets killed or segfaults or similar, this stays -Inf forever and never gets rescheduled?

[mkindahl]

@mindrunner The scheduler will notice that the state is incorrect and correct the next start so that it is rescheduled.

[mindrunner]

Cool, yea I noticed that as well. However, sometimes it seems like nothing is happening at all. Is there some sort of frequency for the scheduler being active? e.g. how long am I supposed to wait after a job failed for it to be re-scheduled?