E@H is not running
16 Apr 2009 9:13:59 UTC
Topic 194294
(moderation:
I had to shut down my computer this morning, and when I powered up again the Boinc Manager reported that I had no tasks lined up for E@H. Furthermore, even if Boinc keeps reporting that it is "resuming computation" my Task Manager tells me that E@H is totally inactive.
Can I re-install E@H?
E@H is not running
)
Yes you can, but it may just be easier to go to the Projects tab and click 'update'. As long as work is available you should get some and then be crunching again. You can also 'detach' from the Project and then 'reattach' sometimes that helps too. It looks like you laptop has work but your other pc is having problems, do you have an anti-virus popping up with problems? If so update it and then try to get some new units. There was a problem with most of the anti-virus programs lately, apparently they all made a change or Boinc made one, and alot of projects units were flagged as problems, but it was all false positives.
RE: RE: I had to shut
)
Well..I did actually install PCTools antivirus program yesterday. But, I have tried reinstalling it, and the other stuff you proposed, but still there are no tasks listed on my Boinc manager, and when I request more work, E@H responds that I already have received my quota??
Anyway, thanks for trying to help. I'm giving this up.
You have returned a lot of
)
You have returned a lot of erroneous tasks, so your quota is set to 5 per core per day (24 hours). This quota was added so rogue computers can't download tasks and return them in error continuously. That doesn't help the science.
So you have to find a fix for the errors you're returning. According to the result files, like this one, you're running into this error:
- exit code -1073741819 (0xc0000005)
Please check this FAQ for what you can do to stop this from happening again.
Everything I that I've
)
Everything I that I've reported is good from what I can see. But I haven't received any new work for seven days now.
I hit the update button. These are my messages:
4/16/2009 8:52:28 PM Einstein@Home Sending scheduler request: Requested by user.
4/16/2009 8:52:28 PM Einstein@Home Requesting new tasks
4/16/2009 8:52:33 PM Einstein@Home Scheduler request completed: got 0 new tasks
4/16/2009 8:53:33 PM Einstein@Home Sending scheduler request: To fetch work.
4/16/2009 8:53:33 PM Einstein@Home Requesting new tasks
4/16/2009 8:53:38 PM Einstein@Home Scheduler request completed: got 0 new tasks
4/16/2009 8:54:38 PM Einstein@Home Sending scheduler request: To fetch work.
4/16/2009 8:54:38 PM Einstein@Home Requesting new tasks
4/16/2009 8:54:43 PM Einstein@Home Scheduler request completed: got 0 new tasks
I'm crunching plenty of S@H, so I'm keeping busy (which is giving me errors that have no relevance here). Heck, I figured I'd just do a reset.
4/16/2009 8:57:30 PM Einstein@Home Resetting project
4/16/2009 8:57:34 PM Einstein@Home Sending scheduler request: To fetch work.
4/16/2009 8:57:34 PM Einstein@Home Requesting new tasks
4/16/2009 8:57:39 PM Einstein@Home Scheduler request completed: got 0 new tasks
I'll be looking at other threads, who knows what is up.
Nevermind. The reset must
)
Nevermind.
The reset must have worked it's magic. I've downloaded some workunits and am now processing.
That just goes to show you... Something.
RE: The reset must have
)
Unfortunately, there's no magic with resetting. All it does is throw away any work in progress that you currently have and throw away the current set of executables for the project and download fresh copies. If the problem wasn't a broken executable or bad data (which it probably wasn't in your case) the reset will actually achieve nothing except a whole lot of unnecessary re-downloading.
What it shows (most probably) is that you weren't allowing for BOINC's use of a debt system in order to make sure that your wishes regarding resource shares are followed as much as possible. If one project (in this case Einstein) does more work than it is entitled to, BOINC will deliberately not download more work until other project(s) have "caught up".
The most likely reason that E@H has had more than its fair share, is a lack of work from Seti around two weeks or so ago. E@H probably supplied extra work when Seti had a shortage. Recently Seti has been "catching up" which is what your list of tasks over on the Seti website seems to suggest. If it really was Seti's turn to catch up, BOINC prevents you from having "stale" E@H work by refusing to download until it is really needed. If you want to understand all this you need to go browse the BOINC Wiki for stuff related to debt, either long term debt (LTD) or short term debt (STD).
One further thing I noticed, which may suggest a possible problem at Seti, is that your list of tasks there shows around 33 tasks waiting to start or "in progress". Quite a significant number of those seem to have been "passed over" in favour of more recent tasks, which suggests that the website has tasks listed (ghost tasks) that actually don't exist on your machine. You should reconcile the list of tasks you actually have against the list the website says you should have and see if there are any discrepancies. If there are, I don't think there's much you can do about it as Seti doesn't use the very useful "Resend lost tasks" feature that Einstein does.
Cheers,
Gary.
What strikes me is that the
)
What strikes me is that the E@H app was apparently requesting stuff
So I'm not so sure this was a debt thing. In older version (I think this is from 6.6.20) boinc used print out something like "requesting xyz seconds of work, reporting N completed tasks", why isn't it doing that anymore?
CU
Bikeman
RE: What strikes me is that
)
His result details from 9 April show he has been using 6.6.20. In this thread, Ageless's first post tells us 6.6.20 has a new CPU/GPU scheduler and improved work fetch policy.
I have 6.6.20 on 1 of my hosts too but it hasn't fetchd any work since I installed it. I have no idea if the less verbose lines are the new way or just an exception. I have heard rumors that the 6.6.20 scheduler and work fetch policy show fairly different behavior compared to previous versions but having had no opportunity to put the thing through its paces I'm not sure whether that translates into "it's bugs" or "it's good".
Anyway, if the 6.6.20 scheduler still uses debts then it seems to me Bryan's story must, in the end, be a debts thing and pretty much the way Gary R. explained it (Einstein received more than its share for some time so it got no work for a while to allow SETI to catch up). Why the coincidence with the reset and the work fetch? I'm guessing the reset simply jiggled the debts in a way that caused the scheduler to download some work. Perhaps only 6.6.20 jiggles that way. Or perhaps it was going to fetch Einstein tasks even if Bryan hadn't done the reset.
BOINC FAQ Service
Official BOINC wiki
Installing BOINC on Linux
Yes, the "less verbose"
)
Yes, the "less verbose" message log is a design choice by the developers.
There are still minor work fetch bugs in BOINC v6.6.20 - this is actually from v6.6.23:
That actually arose because I needed more CUDA work, and my usual supplier (SETI) was offline at the time. But you can see that the verbose debug message actually contradicts the short message shown by default. I've reported this on boinc_alpha.
The flag in cc_config.xml can help track down issues like this.
RE: I have 6.6.20 on 1 of
)
I just read, elsewhere, that 6.6.20 has problems with Intel, and presumably AMD, machines but works just fine on PPC machines. The problem is in getting new work! No more details than that, so no link. Hopefully a new version will fix all this shortly. Some people have said that they are not experiencing any problems, but no details as to to how many projects per machine they are attached too.