Re: FYI: pkg 2.2.1 and official main-amd64 package builder beefy18's time frames: good news and bad news (preliminary)

On Aug 29, 2025, at 12:58, Mark Millard <marklmi@yahoo.com> wrote:

> On Aug 29, 2025, at 11:51, Graham Perrin <grahamperrin@gmail.com> wrote:
> 
>> main-amd64 on beefy18
>> 
>> <https://pkg-status.freebsd.org/beefy18/build.html?mastername=main-amd64-default&build=p9652f95ce8e4_sb45a181a74c>
>> 
>> p9652f95ce8e4_sb45a181a74c began on Tuesday 19th August, I wondered whether it would complete around eleven days later (based on the estimate in <https://lists.freebsd.org/archives/freebsd-pkgbase/2025-August/000741.html>). 
>> 
>> Friday 29th, it's now 75% complete with 8883 jobs remaining.
>> 
>> Complete on Monday 1st September, maybe?
> 
> Well, I expect it goes like this for figuring out a lower
> bound estimate . . .
> 
> ) The overall trend is that the Pkg/Hour rate is decreasing
>  as more builds complete for build of main-* (here main-amd64).
> 
> ) This results in the Pkg/Hour figure shown normally being
>  larger than the final value, continuing to decrease as the
>  overall build sequence progresses.
> 
> ) When I just looked: 8851 Remaining and 108 Pkgs/Hour were shown.
> 
> ) So a likely underestimate would be 8851*(1Hr/108Pkgs)+,
>  So 81.9+ Hours: 3.4+ days
> 
> ) Aug-29 19:12:?? UTC + 3.4+ days is Sep-01 19:12:?? UTC + 0.4+ days,
>  which is about Sep-01 19:12:?? UTC + 10 Hours.
> 
> ) So crudely, still likely an underestimate: Sep-02 05:00:00+ UTC
> 
> But I'll note that the Impulse (the prior 10 min scaled out to
> an hour as if the rate was fixed) seems to be showing figures in
> the 40s through 60s. That would scale out to more like, say, 160
> Hours instead of around 80 Hours more.
> 
> 
> Note: Before the first 10 min have gone by, Impulse scales
> differently --and does not show for the first 2 minutes.
> 

Well, it has been about 44 hours and the 8851 has dropped to 6260,
so somewhat over 58 port-packages per hour, not near 100 per hour.

So, back when I wrote the estimates, the 160 turns out to be more
realistic.

6260/58 approx.= 108 as an estimate from now: so likely over 4 more
days.


===
Mark Millard
marklmi at yahoo.com