Going to 40 percent.
19 Apr 2006 15:32:30 UTC
Topic 191103
(moderation:
Got home from the office yesterday and had yet another power interruption at my residence. 20 plus rigs sitting there useless. Boinc tends to lose its mind under these circumstances and the work (completed or ready to run)evaporates. I spent some time reloading missing work from the server on a few, had to reinstall 40.04 on each, sat down and realized I needed a break. With the cooling season upon us (already), I decided to back off. I will be spending some time retuning some of my crunchers and pulling some apart for housekeeping. I want to thank Akosf for the brilliant work with the optimized code. Too bad the project is clueless as to how to revise server routines to allow a full pull from each contributor. I got a full download of new work on one of my machines last night that was 90 percent 1 hour WU's. The rig blew through each in 13+ minutes and was forced to suck kilowatts with no work until I shut it down this AM. I also want to thank all of you for putting up with my rants about the curse of 32. While recovering work last night I actually was informed by the server that one machine had exceeded its daily quota of 10, and another rig had exceeded its quota of 24....eehhh? Something is amiss somewhere. No fun like debugging ghost legacy code. Hav'nt heard much from the Brucester lately either. I enjoyed being in the top ten and will continue with the project but on a reduced basis.
Regards-tweakster
Going to 40 percent.
)
Your quota gets reduced by one when a WU is sent back errored or times out. This includes WU that time out because your box has thrown away its queue, as you mention here.
As it relates most often to timed out work, my guess is that around a fortnight ago you lost 22 WU from one box, and 8 from another. If I am right you will be able to see those lost WU in the results listed for those two boxes. If this is what happened than it is not a bug, merely a safety feature being overly cautious and working as designed.
The quota is doubled each time you send in a "success" WU, even if that is rejected by the validator. So it would only take two returns to mend the 10 quota, and only one to mend the 24 quota. These three returns would have to happen after the quota is reduced - the fact that you sent in good work in the last fortnight would not help.
The good news is that if you have short WU, you can repair your quota same day by returning one/two of the WU you did get, and then get the balance of the day's quota.
Download 10, return 2+, download another 22.
Download 24, return 1+, download another 8.
The reload can sometimes need a manual update to kick in, though it will often all happen automatically.
River~~
~~gravywavy
Hi Tweakster....Yes, it's the
)
Hi Tweakster....Yes, it's the same situation we had when the 'Alberts' arrived and it took a long time before they finally raised the MDQ to 32. I have the same problem ....one dual core is doing these short WUs in approximately 8 min (thanks to Akosf), does its mixed MDQ of 64 in 14 hours, and must switch to the back up (MalariaControl) for the rest of the day. Good for MC but a 10 hour loss for E@H. This is the second day so I am hoping that this machine will switched to another data set soon. It will be interesting to see if the BOINC client will be smart enough to not process MC@H for a month or so to restore the 95/5% ratio ie. no time loss for E@H over the long term... Cheers, Rog.
I have found that after an
)
I have found that after an unplanned shutdown from power failure etc. that you do not need to lose work. If you set windows to automaticaly run scandisk before windows loads it will fix the files. Setting it this way does not run scandisk every time that you boot the system, just after a failure. The worst that I get from a power failure is rollback to an earlier checkpoint and no lost work.
:)
98SE XP2500+ @ 2.1 GHz Boinc v5.8.8
