Mike,

There's something definitely odd going on here...

I've only received one "new" ARP1 so far, and it is ARP1_0033554_115, a grid cell Mark (MJH333) reported about generation 95 way back in January 2022 - here, _0/_1/_2 all had the 1.5 day deadline I would expect for a genuine Extreme: one of them was User Aborted and I got _3... With such a dearth of first-hand data, I've had a dig into my records and Adri's list...

Looking at some of the rest of the sample Adri provided, there's evidence to back up geophi's observation about generation 126 tasks getting 6 days (so not even being treated as Accelerated?), whereas the generation 125 items cited by Adri got two tasks with 3 day deadlines (as I would expect for Accelerated...).

A random sampling of items in that list for generation 116 and earlier showed up 1.5 days as expected.

It looks to me as if whatever is teeing up the new jobs still thinks that the upper-bound on Accelerated is 125 and [presumably] the upper-bound on Extreme will be 120 (can't tell from this sample, unfortunately!) If my records are correct that would take us back to around 22nd/23rd August 2022, where those values last appeared as bounds in state.txt (Accelerated shifted then, Extreme followed soon after) -- another example of failed communication between the two databases, perhaps?

If that's really happening, it would go a long way to explaining why we are seeing larger than expected average completion times for Extremes and Accelerated tasks in completed.txt :-)

Cheers - Al.

[Edit to slightly re-order some text...]

Al,You got a small miscount there, it should've been three tasks. They ran fine on my device and each one is now waiting for their wingman to complete their copy:

 <1>   ARP1_0017859_125_0  Linux Debian  In Progress           2023-05-18T20:37:17  2023-05-21T20:37:17
 <1> * ARP1_0017859_125_1  Fedora Linux  Pending Validation    2023-05-18T20:37:21  2023-05-19T07:30:45

 <2>   ARP1_0016847_125_0  Linux Ubuntu  In Progress           2023-05-18T20:37:12  2023-05-21T20:37:12
 <2> * ARP1_0016847_125_1  Fedora Linux  Pending Validation    2023-05-18T20:37:21  2023-05-19T06:19:14

 <3>   ARP1_0009060_125_0  LinuxMint     In Progress           2023-05-18T20:37:14  2023-05-21T20:37:14
 <3> * ARP1_0009060_125_1  Fedora Linux  Pending Validation    2023-05-18T20:37:21  2023-05-19T06:19:14

And Mike:Certainly! Like I said earlier, I really should consider only the ones that have less than - I think it was - 10 workunits left, else it will be growing out of hand.

Adri

PS, Validation of older ARP1-tasks is taking place in good (and orderly?) fashion now, last night I had 230 tasks waiting to be validated - with all wingmen's tasks completed and also awaiting validation -, this has shrunk to about 120 over night, after waking up.
EDIT: tags added, I mixed up there and three.
EDIT: added Mike's calculation.

Sorry and thanks, Al, I think I have misunderstood your wording; to me, the word "items" was a blind spot in the sentence. It also would have helped me if you had written "got two tasks each" or "each got two tasks". It's still too early (and it's already past 12 o'clock)! Where is my cup of English tea?

Adri

There's one anomalous outlier in the list I provided (and I've checked generations 099 to 116):
workunit 306934782

ARP1_0031150_109_0  MSWin 11      In Progr.  2023-05-18T20:16:56  2023-05-21T18:06:41
ARP1_0031150_109_1  MSWin 11      In Progr.  2023-05-18T20:17:04  2023-05-20T08:17:04
ARP1_0031150_109_2  MSWin 10      In Progr.  2023-05-18T20:17:06  2023-05-20T08:17:06

That's right, what you see here for task _0 is a deadline of exactly 2 days, 21 hours, 49 minutes and 45 seconds . Why does that device suddenly have a privilege?
(Perhaps some miscalculation popping up. )

Adri
UPDATE:
PS No need to write a new post, just listing some more Extreme workunits, all from generation 126:

ARP1_0004395_126   ARP1_0005710_126   ARP1_0010189_126   ARP1_0011891_126   ARP1_0013933_126   ARP1_0014652_126
ARP1_0004869_126   ARP1_0006292_126   ARP1_0011029_126   ARP1_0012833_126   ARP1_0013937_126   ARP1_0014742_126
ARP1_0005012_126   ARP1_0006692_126   ARP1_0011216_126   ARP1_0013335_126   ARP1_0014161_126   ARP1_0014844_126
ARP1_0005289_126   ARP1_0009833_126   ARP1_0011647_126   ARP1_0013411_126   ARP1_0014310_126   ARP1_0015083_126
ARP1_0005604_126   ARP1_0009995_126   ARP1_0011753_126   ARP1_0013830_126   ARP1_0014373_126   ARP1_0015101_126

The state.txt file for 19th May makes interesting reading! -- it appears to have recovered all bar two of the "missing" units, although the per-classification counts still bear little resemblance to the actual items in each category if we believe generations.txt and then use the "official" definitions of "top ten, next 5, the rest" for Normal, Accelerated and Extreme respectively...

From my notes for the day:

    +--------------+----------------+-------------+
    | number_units | max_generation | type        |
    +--------------+----------------+-------------+
    |           86 | 129            | Extreme     |      was 34;    +52
    |          200 | 134            | Accelerated |      was 57;    +143
    |        35321 | 144            | Normal      |      was 29096; +6225
    +--------------+----------------+-------------+      regained 6420

        Sum: 35607      [only missing 2 now!]

  The actual figures for potential groupings as of this day's
  data are as follows:

    Upper bounds:       129, 134
    Extreme             2357
    Accelerated         3818        [5 moved out]
    Normal             29434        [5 moved in]

I presume that something somewhere in the "automation" for ARP1 task management has finally re-connected to a more up-to-date view of the data; it'll be interesting to see how long it takes it to actually get close to the numbers we might work out from generations.txt, and whether it will increase the number of lower-generation tasks getting 1.5 day or 3 day deadlines in line with the settings current at migration time!

No matter how hard I stare at the numbers in generations.txt for 19th May, I can't get those two totals for Extreme and Accelerated to match up with any possible boundaries; however, the total is close enough to think that at the time it thought Accelerated stopped somewhere around generation 125. That's consistent with the currently observed behaviour of the generation 126 tasks being cited by Adri :-)

Cheers - Al

P.S. speculation: I presume the logistics for ARP1 WU generation management are in the non-BOINC database -- I wonder if there have been problems getting all the validated task information across from the BOINC database?

P.P.S. Adri, I might come back to that thing about strange deadlines, but the above caught my attention :-) [ADHD, no; OCD, perhaps!]

[Edit to fix a transcription error]