HP2 computation errors

[mezcal]

Looking through my errors in more detail I have come a similar conclusion as Xenolith.
Almost all of them errored out in .02 or similar.
Longest was 2 1/2 hours and even that one granted near full points.

I have re-enabled HP2 trusting that they are "on it" but will keep a closer eye on this going forward.

[=[PULSAR]=]

hmmm I wonder if my others errors have something to do with not installing win updates?

[sierra_bound]

I don't think Windows updates have any impact on errors. What I do know, however, is that shutting down or rebooting your computer too quickly can cause errors. This was a problem with Vista which has a tendency to shut down very fast. I have one computer that runs Vista. I always make a point of exiting BOINC first, then waiting ten seconds before rebooting.

[Sparky]

I don't think Windows updates have any impact on errors. What I do know, however, is that shutting down or rebooting your computer too quickly can cause errors. This was a problem with Vista which has a tendency to shut down very fast. I have one computer that runs Vista. I always make a point of exiting BOINC first, then waiting ten seconds before rebooting.

Yeah Vista's fast shutdown would trip up F@H on my GPU all the time as well. At least it usually would restart where it left off but not always.

[=[PULSAR]=]

Is it typical to have a long list of pending validations for a new host?

[sierra_bound]

Yes, that's typical. Until you build up a queue you won't get a steady flow of validations. The reason for that is some work units get validated in a day, while others take a week or more. If you have only a few work units awaiting validation, it may be days before they are approved. You're somewhat at the mercy of the other person who's crunching the identical work unit. If he or she has a very slow machine, it will take longer for them to crunch and return that work unit. Worse yet, if that person has an error, then another identical work unit has to be sent out to another user.

[=[PULSAR]=]

Thanks for the help S_B your always so helpful no wonder your the king

[sierra_bound]

You're welcome. I just checked. I have 29 pages of work units awaiting validation. If you have a lot of crunchers, that's typical.

Ideally, if you have a fast computer, you want the other guy or guys crunching the same work unit to have fast computers too. Unfortunately you're sometimes matched up with a person who's running a slow machine.

[sierra_bound]

One other thing about validations. Completed work units are validated in the order they're received. It's like standing in line outside a movie theater. You start out at the back of the line and gradually make your way to the front.

On weekends, the validation process slows down for some reason and then picks up again when the work week starts. That's one reason why our team numbers tend to jump on Mondays.

[=[PULSAR]=]

If your waiting on other slower computers to complete units doesn't it make sense to keep a higher work buffer, so by the time you start computing someone else with a slower computer already completed it?

[sierra_bound]

I'm not sure about that. There is one advantage to being paired with someone with a slow machine. Sometimes that other person will get a higher claimed credit than you do. If both work units are validated, BOINC awards credit based on an average of the two scores. That will boost the granted credit you're awarded.

The quorum process is sort of like going on a blind date. Sometimes the other person turns out to be drop-dead gorgeous, while other times he or she may be the date from hell.

[=[PULSAR]=]

I'm going to try the 10 day buffer and see how it works out, plus its been really windy the last couple days and would hate for the cable to go out on me.

[sierra_bound]

I think 10 days is a little too much. 4-5 days should be enough unless you're anticipating a prolonged cable outage. The wind may end up knocking out your power as well.

[fallwind]

I haven't found much benefit in running a large cache in terms or reducing my pending validation queue. Once your machine goes over 2 days for it's average turnaround time, it is disqualified from receiving make-up a or repair WU's, ie- WU's that were sent out earlier but not returned or returned with an error or invalid. These WU's are validated and points awarded immediately when you return them which helps reduce your PV queue. Another way to reduce PV is to run a zero redundance project like FAAH, here you will never have a PV queue, only a few that are marked "inconclusive" and sent out to a second computer as a safety check on the work your machine sent in.

[xenolith]

I think 10 days is a little too much. 4-5 days should be enough unless you're anticipating a prolonged cable outage. The wind may end up knocking out your power as well.

Who lost their power today?

<:::raises hand:::

I had no crunching for five hours due to 60mph winds.

[T_M]

30 errors for this project since the 20th
All failed very quickly though so not sure i will bother dropping it.

[Chris_redfield]

I haven't been paying much attention recently but only just noticed this. Had a quick look at my stats over on wcg and human protene folding has been churning out errors

Pretty sure its not my system as its not OC'd at all.

[OldChap]

I believe this is an OS problem rather than an OC problem...Vista and 7 are prone to errors

[Wolfsburg18]

Thanks for the information I just looked at my stats
63 Tasks Run
9 in Progress
4 Valid
50 Error (Wow)

In some cases the error tasks took 15 or more hours of CPU time. Roughly 2k of Wasted CPU time.

[Wolfsburg18]

I believe this is an OS problem rather than an OC problem...Vista and 7 are prone to errors

I would counter this because I am running Linux on 3 crunchers that have generated errors. Only my primary computer is running Vista 64 and results from it have about the same mixture of being valid vs error status.

[xenolith]

I would counter this because I am running Linux on 3 crunchers that have generated errors. Only my primary computer is running Vista 64 and results from it have about the same mixture of being valid vs error status.

By what you're describing, I would say the version Linux you're using is by far worse than any other OS.

50 returned error WUs totaling 2000 CPU hours? That can't be right.

[Vapor]

If you're getting any errors, it's definitely worth dropping! The erroneous units may only report using a few minutes of CPU time, but I know my CPU hours on a fully-dedicated quad core were around 56hrs/day when it was enabled. Disabled it and averaging 95-96hrs/day now

[D_A]

By what you're describing, I would say the version Linux you're using is by far worse than any other OS.

50 returned error WUs totaling 2000 CPU hours? That can't be right.

It's the units, not the OS. If you re-read his post he says the Vista is getting the same mixture of errors and valids.

[xenolith]

It's the units, not the OS. If you re-read his post he says the Vista is getting the same mixture of errors and valids.

I understand it's the WUs not the OS. But by his initial description of 50 WU errors wasting 2000 CPU hours, that by far is the worst I've seen, regardless of OS. (I've been looking around at a few places concerning this common error.)

Yesterday, I dropped HP2 just to be done with it...

[D_A]

I suffered through the errors long enough to get my badge, but I won't touch it after that. Too many errors, many many reports and the HPF2 team still won't do anything about it. By comparison, FAAH seems almost bullet proof. Can't kill it with a stick.