| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Daily Quota Strangeness |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 12
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Along with (and probably as a result of) the current problems with HPF2 units, I've noticed some issues with the Daily Quota not working as I thought it was supposed to.
It's my understanding from googling that the Daily Quota is supposed to be reduced by 1 (per CPU) whenever an Error result is returned, or no result is returned by the deadline. And that the Daily Quota is supposed double (up to some project-determined limit) when a good result is returned. The result of this being that the Daily Quota should be generous enough to stay out of the way and not become an issue unless you return a bunch of Errors in a row with no or few offsetting good results, thus keeping broken systems from screwing things up too much. But on several days since HPF2 started, towards the end of the UTC day, I've had new work downloads denied with "No work sent, reached daily quota of (4 or 8) results". Often, I can get around this by waiting for a result or two to finish and then forcing a manual update to report them, increasing my daily limit and allowing more work to be fetched. But if I don't happen to notice it in time, my system will run out of work and sit idle for a few hours until BOINC tries again the next UTC day. I don't understand why the Daily Quota is limiting me. I usually turn in around 12-15 results a day on my AMD X2 machine, and even with the current HPF2 problems, I've never seen more than 1-2 Errors in a given day, nor have I failed to return any results. With 1 Error and 11+ good results (or 1 Error and 5+ good results per CPU, if that's how it works), the Daily Quota should hardly twitch off of the maximum allowed. How is it getting reduced to 2/4/8 or whatever? It almost seems like the project must be decreasing the limit for more than just Errors, or else is resetting it much more drastically than expected for each Error. Although I do still get a few Invalids per day (discussed elsewhere), even those aren't enough to explain the limits. It would have to count every Inconclusive as an error as well, which it obviously shouldn't. So, how's it supposed to work? And is there anyplace where I can see my current Daily Quota (other than the error messages when it kicks in)? I haven't been able to find one. Oh, and it's BOINC User ID 225561, Host ID 41341, in case someone wants to look into this. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
RickH, I think 1 error is enough 'per device' to make it kick in a daily quota of 2. For each 'valid' returned, after that there might be a safety to let you have one at the time, when expressly asked thru the update mechanism???
----------------------------------------There's an 'uhhh wonderfull' thread, that was started before this one that discusses same. Also remember knreed writing something about it a week or so ago...send mail to support and he can reset, but if you keep on sending 1 or 2 errors a day, even if you do 6 per device, because that's how it functions, it will kick it back in immediately..... But why 1 or 2 errors a day....on a good cruncher, errors should be an exception. Were you not writing something about a rock solid test elsewhere and the AMD socket 939 CPU u have and the foobar instruction it has in the 15th decimal? No errors, no quota limitations.....i think you can send as many inconclusives, pending validations, invalids as you can from 1 computer, but errors means 'The Device' failed to meet the basic set of 'pre-validation' conditions on a finished WU or prematurely ended it, intermittent as it may be. I think WCG techs need to advise on the dilemma with the FUBAR instruction.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The Error results are clearly an HPF2 problem. When I get one, it's not just me; the other copies of the WU also end up returning errors. For example, the one I just reported earlier today:
za097_ 01248 Error 07/11/2006 08:52:44 07/11/2006 11:44:53 1.47 17 / 0 za097_ 01248 Error 07/11/2006 07:16:44 07/11/2006 12:52:15 3.31 17 / 0 za097_ 01248 Error 07/11/2006 06:59:55 07/11/2006 08:51:42 1.80 19 / 0 za097_ 01248 Error 07/11/2006 02:43:37 07/11/2006 07:15:26 2.86 17 / 0 za097_ 01248 In Progress 07/11/2006 02:43:13 07/18/2006 02:43:13 0.00 0 / 0 za097_ 01248 Error 07/11/2006 02:41:07 07/11/2006 06:55:34 1.17 22 / 0 The reason I see so many is that I crunch so many units. Don't think of it as 1 a day; think of it as 1 or 2 out of every 30 WUs or so. The other thread you're referring to was talking about the large number of Invalids I'm also seeing, which is happening to some other people but not many, apparently. I'm not worried about why I see Errors; it's a general HPF2 problem that the techs are busily working on. And I know the Daily Quota is supposed to get knocked back when you have an Error. My specific question was why is it getting knocked back so far so fast? Everything I've read says it shouldn't get knocked back to 4 for a single Error, so I was checking if we've got another bug going on (that we haven't seen before because you wouldn't normally get enough Errors to notice it). Maybe this project is set to knock your quota all the way back to 1 or 2 on every Error. If so, that's fine. I'm just asking, because other things I've read have clearly stated that the normal rule is "down 1 per Error, double on good." |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi RickH,
I do not have any answers. Here is the way it is supposed to work: http://boinc-wiki.ath.cx/index.php?title=Daily_Quota Maybe we have a bug in the new server code? Lawrence |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yeah, that's what I thought. Unless there's a whole string of errors, the quota should stay plenty high enough to keep out of the way. And I'm not getting strings of errors, just one a day or so. The only non-normal thing I'm getting enough of to have any possible effect would be if it's counting Inconclusives as errors, which it shouldn't.
Oh, well... something else for the techs to chew on, as if they didn't have enough to work on already.  |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Maybe reading this Wiki back to front would illicit me on posting many 'corrective suggestions' on their message board......says that the number for WCG is 'unknown'!!!!!!
----------------------------------------If RickH is online 24/7 would a lower contact period reset to the default 2.4 hours e.g. ** not automate the retrieval entirely upon return of a good result.....like you've been a good boy/girl, give you one more, but not a truckload (like described in the wiki)? His number of Pending Validation would increase, given he'd always have WU's freshly send to 3 within a few minutes. ** Some set their contact Temporarily to e.g. 5 days, go offline, travel, return update....changing the applicable profile is done in a jiffy. 3 Profiles can be set at WCG with fixed names home, work, school....i think of school as my travel profile... any profile gets copied back to the client upon hitting the update button in BOINC.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Jul 12, 2006 8:51:30 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hmm, I don't really get what you mean here. My contact period is set to the default 2.4 hours (0.1 days), since this machine is connected 24/7.
I normally have 2 WUs being crunched, with 2 more queued up ready to go. Within a few minutes after I finish one and start on the next, it reports the recently uploaded finished one and downloads another WU so that I have 2 queued up again. The problem is that around 20:00-22:00 UTC, when it tries to get the next, it'll probably tell me I've hit my quota again and refuse to fetch it for me. I never get truckloads; a finished one goes out, a fresh new one comes in, which is what I'd expect when you're always connected. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Don't rely on the wiki for the policy details - it may not be correct.
I'm afraid you can't rely on the documentation at all, since WCG don't use the bleeding edge BOINC server release, and the version they do use is customised. So, we will have to wait for an official pronouncement. However, I do recall knreed telling us he had reset the quota to 20 for everybody. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Think the C.A.'s should have a private forum for consultation, agree and than help us ignoramus (who feel to be in that category regarding the WCG/UD/BOINC domain)....one referring us to the wiki, the next shooting it down.
---------------------------------------- RickH, if its already at 2.4 hours, i'm afraid you'll have to life with an every 6th error......each generated requiring a 4th copy to be send....no idea how many there are with this FUBAR instruction being responsible as you indicated, that would motivate for a WCG tweak in the HPF2 crunch logic.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
We do, thank you. However, we don't consult over every trivial answer we give, and I'm not contradicting Lawrence - merely down-playing the wiki a little. It's still very useful information, for those interested in how the quota works.
|
||
|
|
|