Hmmm..... I still cannot see any update on BoincStats for MW@H. I cleared my browser and reloaded the site from scratch and still blank for MW. Maybe it will sort itself out by tomorrow.

Cheers, Keith

---

Just testing mine too!
Seems pretty close.
Now if we could just get WCG to fix their stats things would look correct!

---

I still cannot see any update on BoincStats for MW@H

You will not see an overall total until BOINCStats does the daily update and consolidation at 3pm UTC. You will see the interim updates by checking your individual all up total page for MW (you got 11,350.14 coming to you on the daily update)

Thanks Zydor, I checked just now and see the daily update for MW finally took and I'm getting daily credit again. Now for the other problem that's been bothering me lately with MW. I used to average somewhere around 50-60K daily RAC and for some reason, my RAC has plummeted to only 3K. I hadn't changed anything on my two crunchers, yet I never could see them doing MW work. Einstein seems to have taken priority for the last couple months, yet it has the lowest resource share. I just increased the MW resource share again and hope to see Einstein@Home take a backseat to Seti@Home and MilkyWay@Home. Anybody else seeing their RAC for MW take a dive over the last couple months and anyone have an explanation?? Since the server went down only for a couple of weeks, I can't see how that would have affected my RAC so drastically. I backleveled to 6.10.60 Manager about the same time in order to avoid the idiotic backoff of the latest version and to get enough Seti work and that has returned me to proper levels of work for E@H and Seti@Home. Does the 6.10.60 Manager not correctly implement the resource share you set for your project preferences?

Cheers, Keith

---

Anybody else seeing their RAC for MW take a dive over the last couple months and anyone have an explanation?? Since the server went down only for a couple of weeks, I can't see how that would have affected my RAC so drastically.

The RAC is falling extremly fast when such problems like MW had appear. I Have a (~30k) 24/7h machine for MW with no GPU Backup projects on it, so i get every WU that i can become the latest months. So when you use backup projects, it can be, that i run several WUs (plus refresh the cache) while you compute "in this one working hour" a backupproject. When there was absolutly nothing to compute i falling too like hell. I had the luck my unattended MW Machine has no problems with the serverchanges (IP, DNS Changes etc.) at all.

---

DSKAG Austria Research Team: http://www.research.dskag.at

There is something else strange going on with MW on my crunchers. I have the full allotment of tasks on board each machine, currently 70 per machine, but each machine only processed 4 tasks per GPU in a 24 hour period at a recurring time in the day for the last couple of days. Each machine has two GPU's. If I let all projects run, then MW only does work once a day. The only way to get MW to process WU's is to suspend my other two projects. The machines are not obeying the resource share set in preferences. MW should run 33% of the time, Seti 55% of the time and Einstein 11% of the time. Instead, only Seti and Einstein are cycling through the resource share each hour which is what is set in preferences.

Does anyone have an explanation for the behavior I am seeing?

Cheers, Keith

---

There is something else strange going on with MW on my crunchers. I have the full allotment of tasks on board each machine, currently 70 per machine, but each machine only processed 4 tasks per GPU in a 24 hour period at a recurring time in the day for the last couple of days. Each machine has two GPU's. If I let all projects run, then MW only does work once a day. The only way to get MW to process WU's is to suspend my other two projects. The machines are not obeying the resource share set in preferences. MW should run 33% of the time, Seti 55% of the time and Einstein 11% of the time. Instead, only Seti and Einstein are cycling through the resource share each hour which is what is set in preferences.

Does anyone have an explanation for the behavior I am seeing?

Cheers, Keith

Do you 'owe' time to the other projects? What is your long term debt on those other projects? Another way to 'fix it' would be thru a cc_config.xml file, I think, that excludes 1 of your cards from anything but MW work. Of course then when MW is out of work so is your card!

Thanks for the reply Mikey. From my reading up on long-term debt, the sums of all project long-term debts add up to zero. My E@H debt is -86K, the Seti debt is +21K and MW@H debt is +56K. I am assuming that GPU scheduling priority = long-term debt. Those values roughly add up to zero. I have made an assumption that the large E@H debt is because of a large cache built up. I looked at my project preferences and see that the project seems to have inherited my 4 day cache setting I have made primarily for Seti. I never worried about enough work for E@H nor MW@H since Einstein seems to be downloading constantly and with the work scheduling system for MilkyWay a kind of just in time system where it always maintains the same number of work units on board. I have adjusted the work cache for Einstein to zero in the hopes of retiring such a large negative long-term debt and will wait and see whether MilkyWay ever starts processing its largish positive long-term debt.

Keith

---

If it gets real messy, and it sure can at times in that kind of scenario - the "get out of jail free" card is the following line inside cc_config in the BOINC directory

<cc_config>
<options>

<zero_debts>0/1</zero_debts>

</options>
</cc_config>

That will reset all debts to zero every time you start BOINC until its removed or the value set to zero not one from cc_config.

"1" switchs on the zeroing out of debt, "0" nothing happens

Regards
Zy

Thanks for the reply Mikey. From my reading up on long-term debt, the sums of all project long-term debts add up to zero. My E@H debt is -86K, the Seti debt is +21K and MW@H debt is +56K. I am assuming that GPU scheduling priority = long-term debt. Those values roughly add up to zero. I have made an assumption that the large E@H debt is because of a large cache built up. I looked at my project preferences and see that the project seems to have inherited my 4 day cache setting I have made primarily for Seti. I never worried about enough work for E@H nor MW@H since Einstein seems to be downloading constantly and with the work scheduling system for MilkyWay a kind of just in time system where it always maintains the same number of work units on board. I have adjusted the work cache for Einstein to zero in the hopes of retiring such a large negative long-term debt and will wait and see whether MilkyWay ever starts processing its largish positive long-term debt.

Keith

Long Term Debt, LTD, is accumulated when one project has very little work and your pc crunches for other projects instead, therefore Boinc 'owes' time to the project that is currently no providing work to your pc. The debt is set by the percentage setting you use for each project, like I crunch for Rosetta and Malaria on this laptop, 80% Rosetta and 20% Malaria. When either has no work when I need it the other project gets more work and Boinc then 'owes' time to the other project. It can be a VERY long term process to even out the debt so alot of people use Zydor's cc_config option to just zero it out and move on. Me I change mine manually but it is the same thing, I chose the projects I want to crunch for and how much time to spend on each. I am interested in short term times not long term times, so I manually reduce mine when it goes wonky! I do not however change mine to zero, I just go in and change the number to something like 1.48868809 or whatever the numbers are after the decimal point, which is close enough for me.

LTD builds up for BOTH cpu AND gpu projects, the newer versions of Boinc are better at separating the two, the old versions just treated them as one project no matter whether it the cpu or the gpu. It is a work in progress for Boinc and the programmers.

Thanks for the explanation, Mikey. Well, I got impatient and tried Zydor's cc_config trick and zeroed the long term debt. I then got all projects into a high-priority mode for a little while before things settled out. I am seeing MW running on one computer now for what looks like the correct sharing interval with my other projects. I haven't seen any sign of MW running on the other computer however. That computer also is showing something like 8 hours of work fetch deferral with 17 hours of work fetch interval. The other computer that ran MW early this morning has what looks like a reasonable set of values for those parameters. So, again impatience won out and I reset those parameters for that computer to zero and will wait and see what happens. Dont' know what is going on with that computer as my comparison between the two computers essentially shows the same values. One computer is now running MW correctly and the other is still a no-show.

Cheers, Keith

---

Thanks for the explanation, Mikey. Well, I got impatient and tried Zydor's cc_config trick and zeroed the long term debt. I then got all projects into a high-priority mode for a little while before things settled out. I am seeing MW running on one computer now for what looks like the correct sharing interval with my other projects. I haven't seen any sign of MW running on the other computer however. That computer also is showing something like 8 hours of work fetch deferral with 17 hours of work fetch interval. The other computer that ran MW early this morning has what looks like a reasonable set of values for those parameters. So, again impatience won out and I reset those parameters for that computer to zero and will wait and see what happens. Dont' know what is going on with that computer as my comparison between the two computers essentially shows the same values. One computer is now running MW correctly and the other is still a no-show.

Cheers, Keith

I hate to say this but Boinc can take more than a MONTH to settle down and crunch right when you crunch multiple projects on one pc, which is the whole reason I have multiple pc's! Go into that last pc and check ALL your settings, something is different, projects being run, cache, pc hardware...something!