Trippple Checks

[GP2]

Originally Posted by Madpoo

Odds are they need to restart the program and maybe it would finish? Just a guess. We'll find out eventually...

But it keeps checking in every day, so the program has to be running still.

As of now: "Updated" = 2017-06-06. Tomorrow it will be 2017-06-07.

Either they edited their worktodo.txt to move it down below something else (any other active assignments on this computer making progress?), or maybe they now only turn their computer on for five minutes each day and it's doing 0.01% a day or something.

At least we know it's below 99.5%, because if it was higher than that then the exponent status would show "% Done" = 100.0%, yet still not completed. Yes, I've actually seen that displayed before... in that particular case, the exponent completed normally very shortly afterward.

[Madpoo]

Originally Posted by GP2

But it keeps checking in every day, so the program has to be running still.

As of now: "Updated" = 2017-06-06. Tomorrow it will be 2017-06-07.

Either they edited their worktodo.txt to move it down below something else (any other active assignments on this computer making progress?), or maybe they now only turn their computer on for five minutes each day and it's doing 0.01% a day or something.

At least we know it's below 99.5%, because if it was higher than that then the exponent status would show "% Done" = 100.0%, yet still not completed. Yes, I've actually seen that displayed before... in that particular case, the exponent completed normally very shortly afterward.

That CPU doesn't have any other work assigned and hasn't turned in anything else recently. Yeah, it just seems like something is mucking it up. Maybe something else on there started pegging the CPU at 100% a month ago leaving Prime95 precious few spare cycles. Or... who knows.

It's running this, fyi: Windows64,Prime95,v28.10,build 1

EDIT: You thought that one was bad, how about this one? 75318277. Or this DC: 41385041

[Dubslow]

Haven't we heard reports of somehow Prime95 forgetting it has work assigned, thus acquiring new work before somehow finding the old work again and appending it to the worktodo file, after the newly acquired work? Which would almost certainly lead to a situation such as this if the timing was extraordinarily unfortunate.

Edit: Crosspost, whoops. Apparently this CPU has nothing else assigned so there goes that theory

[LaurV]

Something is fishy with all those 99.9%. Maybe there is a bug with reporting (either on P95 client or on the server side), just a wild ass guess... It smells like a reporting issue... Maybe the user just finished the work and reported it, but for some strange reason it was not properly sent or recorded by the server. Can we have a situation of how many expos are in such situation, like blocked somewhere over 90% for a long time? (it may be that the last report before submission was not 99.9%, but a bit lower, some people do not report daily, but report weekly or monthly, etc).

[Madpoo]

Originally Posted by LaurV

Something is fishy with all those 99.9%. Maybe there is a bug with reporting (either on P95 client or on the server side), just a wild ass guess... It smells like a reporting issue... Maybe the user just finished the work and reported it, but for some strange reason it was not properly sent or recorded by the server. Can we have a situation of how many expos are in such situation, like blocked somewhere over 90% for a long time? (it may be that the last report before submission was not 99.9%, but a bit lower, some people do not report daily, but report weekly or monthly, etc).

The 2 extra ones I mentioned are actually in a slightly different category... yeah, they've been at 99.9% a while but not nearly as long as that first one that started this.

In those 2 cases I think the machine is just extraordinarily slow and only progresses a fraction of a % in any given week. What's remarkable for those is that they're even still running and checking in, but at least with them I have a hunch they'll probably finish in another several months.

That first one though, it's really strange. I don't know what to say except the person would probably do well to restart Prime95 in case something hung up internally. If they stopped the LL test but left Prime95 running, is that something where it would still check in daily even without any workers going? Maybe he stopped the worker to do something (play a game or whatever) and was coincidentally at 99.9%, then he forgot to start it back up. Out of sight, out of mind.

[S485122]

Originally Posted by Madpoo

...
If they stopped the LL test but left Prime95 running, is that something where it would still check in daily even without any workers going? Maybe he stopped the worker to do something (play a game or whatever) and was coincidentally at 99.9%, then he forgot to start it back up. Out of sight, out of mind.

Not sure about a stopped Prime95 instance checking in with PrimeNet. But one thing is sure : after rebooting the computer Prime95 would start again and not in paused mode. And while there are computers that did not have to restart in three years, it would be a coincidence if this was one of those.
It might be a case of a wrong "PauseWhileRunning=" clause in prime.txt.

Jacob

[Madpoo]

Since I capture a daily snapshot of the progress of assignments, I was able to generate this little gem of a report. Any assignment that has reported 99.9% done more than once.

In descending order of "stuckness" (how many times it reported 99.9% over the total # of datapoints I've collected for it):

exponent	%done	Stuck	DataPoints
42702343	99.9	15	17
42585419	99.9	3	4
75976063	99.9	47	66
43026001	99.9	2	3
44216219	99.9	38	59
80398193	99.9	20	46
45165577	99.9	17	74
42782879	99.9	5	28
79374497	99.9	18	129
79661359	99.9	9	118
79694897	99.9	5	83
77492189	99.9	2	57
79964207	99.9	2	62
77422633	99.9	2	84
43545401	99.9	3	149
41385041	99.9	2	222

[kladner]

That sounds interesting. I'll take the first one
42702343

[GP2]

Originally Posted by Madpoo

Since I capture a daily snapshot of the progress of assignments, I was able to generate this little gem of a report. Any assignment that has reported 99.9% done more than once.

Might be worth a quick check to see if there are any stuck at 100.0%

I have seen 100.0% reported once before, although that was for an exponent which completed very shortly afterward.

Edit: never mind, 79964207 from your list is currently reporting 100.0%

PS, actually the list might be partly out of date, for instance 42782879 is at 23.7% and 45165577 at 55.1%

PPS, lots of Xolotl exponents in there, can he be contacted for insight into what's happening here?

[Madpoo]

Originally Posted by GP2

Might be worth a quick check to see if there are any stuck at 100.0%

I have seen 100.0% reported once before, although that was for an exponent which completed very shortly afterward.

Edit: never mind, 79964207 from your list is currently reporting 100.0%

PS, actually the list might be partly out of date, for instance 42782879 is at 23.7% and 45165577 at 55.1%

PPS, lots of Xolotl exponents in there, can he be contacted for insight into what's happening here?

There were some weird things in there when I looked closer.

These exponents, for example, actually seemed to start over (same assignment). The 99.9% was the highest they got and then they either rolled severely back (restored a temp file?) or started over entirely.
44216219
45165577
42782879
77492189

This exponent is stuck at 100% now: 79694897

These exponents are stuck at 100%:

exponent	Stuck	DataPoints
77494063	41	52
42448387	7	11
80019941	20	47
77263517	29	152
79828501	8	68
46216259	3	28
79694897	5	83

One of those, 79828501, rolled back at some point and is now 73.6% (although as a note, none of them have updated recently anyway).

[kladner]

Originally Posted by kladner

That sounds interesting. I'll take the first one
42702343

I matched c cooper. Xolotl has yet to finish.