Where are the tasks?
9 Mar 2009 23:06:23 UTC
Topic 194227
(moderation:
I recently upgraded my computer (moved from a PPC Mac Mini to the new iMac).
I transferred all my files to my new computer and I've seen an immense increase in the processing speed for boinc apps. (I run SETI and Einstein). (2-4 hours for Einstein from 4-8 days).
Even before the transfer, though, I'd noticed that it would often take 2-3 days to get new Einstein tasks. I would even try updating but it wouldn't request any new work.
I'm running the Boinc 6.2.18 with a SETI optimizer for multibeam and astropulse, but the latter is new as of the new computer and the issue is pre-existing.
From the other similar topic "no more tasks", I don't know how I could be running a beta test app as I'm not sure what one is in reference to boinc.
So, now SETI tasks are getting all my time as it regularly sends me a queue of tasks to process even before I finish what I've got already. However, when I do have tasks for Einstein, it seems to use one processor for each.
Where are the tasks?
)
Check in he XML for long term debt values. If einstien was monopolizing your time in the past, boinc will schedule seti almost exclusively for a while to catch it up.
Alternately, what're your resources shares and uptime percentage? If your mac's almost never on it might not be requesting work because it's concerned about finishing your seto WU's before they time out.
RE: I recently upgraded my
)
How long did it take to do the transition from the old machine to the new? If BOINC was off for a couple of days, it will have noted this and will therefore assume that in future it may not get access to your machine in a timely manner. This will cause it to be frugal with the quantity of work it downloads until you get back to your normal pattern.
Have a look in your state file (client_state.xml) and see what values are recorded there for the parameters and . Just browse the file with a text editor and don't make any changes. Also it would be helpful if you tell us the resource shares you have allocated to the two projects. Are you heavily weighted towards Seti?
How many hours a day do you normally run your machine and do you allow tasks to be crunched while you are using your machine for other purposes?
Cheers,
Gary.
My computer's never off and
)
My computer's never off and Boinc is set to run at all times, even when the user is active. It's set to use up to 50% when the user is active and 90% when idle. Also, the computer is predominantly idle, as I use it less than 2-3 hours per day.
I'm sure there are a lot of people who don't run the program often, but with such a minor draw on electricity, I find no reason to ever turn my computers off. Additionally, Boinc is set to auto-restart whenever my computer comes on if I do happen to restart (for updating or whatever).
I'm not sure which XML file to look in for the long term debt values. I found one in the same file mentioned in Gary's post and it's -942532.310523 for Einstein and the same but positive value for SETI.
I have Boinc set to switch between projects every hour (not so much necessary with the dual core). The projects have an equal share of time (50/50).
The transfer took around 6 hours or so and I have had a Einstein project since then. Infact, when I updated to get the system to recognize the computer upgrade, I got a new task and it was completed in 9 hours.
The numbers in the state file mentioned were around .96 for the on and .81 for the active. If my guess is correct, these numbers are actually the percentage of time that Boinc has been on (for the first) out of all the time my computer's had the program and the percentage of time it's been able to run (for the second)? If so, the lower number for the latter is due to the fact that, on my old computer, it was impractical to allow Boinc to crunch numbers while I was using it as the system was extremely slow.
Thanks to both of you for your assistance.
~Catherine
RE: My computer's never off
)
That would be percentage of memory usage, since you can't adjust CPU usage separately for active and idle (and shouldn't throttle anyway).
You took the right file, and obviously DanNeely hit the point: Einstein is in SETI's debt for about 11 days of computing time. So BOINC will prefer SETI for a time, until each project has indeed had its share of 50%.
I think your analysis is quite correct, and the numbers are pretty normal.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: I'm not sure which XML
)
As Gundolf has mentioned, Seti is owed about 11 full days of crunching and this is why BOINC is not wanting to download E@H tasks. This implies that there was a period of time where you must have crunched E@H for many days - perhaps as a result of Seti not being able to supply work. Seti occasionally cannot supply work but it seems strange that the discrepancy should build to 11 days. I'm more inclined to think that the debt value has become corrupted in some way.
If you want to quickly get back on a more even keel, you can "fix" things very easily. Just stop BOINC and open the state file (client_state.xml) in a simple text editor (like Windows notepad) that doesn't mess with the formatting of the file. Simply remove the first 4 digits of each debt value. The E@H debt would thus become -32.310523 and the Seti debt would become 32.310523. Now Seti is only owed 32 seconds of debt so both projects should be able to get work and be equally active. The actual number of seconds is not important - you could completely zero the debts if you wanted to. It's just easier to knock off a couple of leading digits which is why I suggested it. Whatever you do, you must keep the positive and negative values precisely the same (just as shown in the example). Actually, more precisely, the debt values must sum to zero. Once you have made the edit, save the file and restart BOINC.
I'm not trying to nitpick but the on_frac is the fraction of time your machine is on. 0.96 means that your machine runs approximately 23 hours per day on average. The active_frac is the fraction of time that BOINC is active while your machine is on. 0.81 means that BOINC is active for about 18.6 hours out of the 23 hours per day that your machine is on. If you left your machine on solidly, without interruption, for the next month or so, both of these numbers would rise and approach 0.999+ over that time.
As Gundolf mentions, these numbers are quite normal and it's only the debt values that seem to be highly suspect. If it were my machine, I'd be "fixing" those debt values ;-).
Cheers,
Gary.
Will do. Thanks so much for
)
Will do. Thanks so much for the help.
~Catherine
RE: Will do. Thanks so
)
No problem! Thanks very much for your contribution to the project.
PS: I see you have fresh tasks now so all looks fine.
Cheers,
Gary.
It seems that the two tasks I
)
It seems that the two tasks I managed to pull in were a temporary aberration. I've altered the debt twice but each time I restart Boinc, it goes back to what it was. I guess I'll just have to wait. At least it's decreasing. Right now it's down to 690K and change. If I reset it using the "reset" button would it do anything beneficial?
RE: It seems that the two
)
Doing a Reset on a project that has a large Debt (positive or negative) will bring the value about half way to zero). DO NOT do a reset if you have any WUs in the queue for that project, they will be thrown away by BOINC but NOT reported back to the project as being gone. A Detach is better for a case like that, however, if you have any optimized apps for that project they will need to be reinstalled.
[edit]If you're still having problems getting work for a project you need to check your cache settings. If you have an always-on connection (cable/DSL), I recommend setting the Connect Interval to 0.1 and the Additional Work to how many days you want work for (up to 10 days). If you have dial-up then you need to decide how often you want/can allow BOINC to phone home and set the Connect Interval accordingly.
If you set your Connect Interval too high, you can cause a project with short deadlines (some SETI MB WUs have short deadlines) to either NOT get work, or if they do get work they will go into high priority mode (used to be called EDF-earliest deadline first) until they are crunched. If your system goes into high priority mode it will not allow any downloads until they are cleared. A worst-case instance of this can result in a quad-core system running only ONE CPU and the cache being empty.[/edit]
Seti Classic Final Total: 11446 WU.
Since the problem is that I'm
)
Since the problem is that I'm not getting any WUs (and so don't have any at all), I'll try resetting first. I don't like having caches for several days in the future because I think it's silly to have a task sitting on my computer and not being processed for several days when someone else could actually be actively processing it.
I wasn't aware that there was an optimization program for Einstein. I have two or three of them for SETI but none for E.
~C