no new work because???
17 Jul 2009 22:21:56 UTC
Topic 194446
(moderation:
What does this last sentence mean? I quote:
17-7-2009 22:38:25 Einstein@Home Scheduler request completed: got 0 new tasks
17-7-2009 22:38:25 Einstein@Home Message from server: No work sent
17-7-2009 22:38:25 Einstein@Home Message from server: (won't finish in time) BOINC runs 42.7% of time, computation enabled 100.0% of that
no new work because???
)
A possible bug in BOINC clients later than v6.6.31
It's being investigated.
RE: What does this last
)
It could also mean you have several projects running at the same time on your pc and Boinc now thinks it can't download a task from Einstein and finish it in time since your other projects have deadlines too. It could also mean you have Boinc set to only run part of the day, not 24/7, and in conjunction with the above sentence won't finish any work downloaded before the deadline due to time constraints.
Well, the scheduler went to
)
Well, the scheduler went to great length trying to find suitable work:
From the scheduler logs:
From what I get, this seems to be the crucial line:
which is info sent by the client. Note the "delay" of 6 million seconds or ca 70 days ... The scheduler then tries several candidate jobs and always concludes that the "delay" plus the estimated runtime of the job exceeds the remaining time to the deadline ... and finally gives up.
So yes, it's a problem of the client that asks for new jobs but at the same time tells the server that the work will not be finished in the next three months...
Under Linux, I've downgraded from 6.6.x to 6.4.5. again. Not sure why for Windows 6.6.x is now recommended while under Linux it's still considered an unstable development version. I've seen so many odd scheduling effects in 6.6.x (which should be identical for Windows and Linux as it's not OS specific) that I don't feel comfortable running 6.6.x, x <= 36
CU
Bikeman
RE: From what I get, this
)
There's been some progress on the development front. Other people have homed in on that "delay" figure: the first attempt at fixing it made things worse, but a later one is reported to be better. Now all we need is a compiled v6.6.38, and then we can test it.....
I've built the new client
)
I've built the new client from source and will test it. Got one machine on Malaria complaining about the same things, so it's a good testbed there.
RE: Under Linux, I've
)
I am using 6.6.31, which declares to be 6.6.29 on 5 BOINC projects. The only one which gives problems is CPDN Beta.
Tullio
RE: I am using 6.6.31,
)
I run 6.6.31 too on Linux (Fedora 11 x86_64) and mine reports 6.6.31 on all projects (7) I'm attached to (Einstein task). Difference might be I run 64bit BOINC and you seem to run 32bit... Also haven't seen other reports (well, not from other ppl) about this. Are you sure your core client got updated to 6.6.31? I simply extract files from the archive downloaded from BOINC-site, then copy the 3 main files to correct folder, replacing older files.
I've been tearing out hair
)
I've been tearing out hair over several problems that made no sense,including BOINC asking Einstein for work for the GPU (not the CPU) which was resulting in no more work for Einstein. I couldn't find any place where I had requested GPU work for Einstein, since AFAIK Einstein didn't use CUDA. In fact, I couldn't find any place that I COULD have specified such an option. I was using CUDA on SETI, which is in a pretty sorry state these days and confusing the issue because it can't receive results in a timely fashion. Their system seems badly overloaded.
On a hunch, after reading the messages on this thread, I removed the latest version of BOINC (installed recently) and installed 6.4.7, the most recent earlier version that I could find on the BOINC site. Within minutes it was clear the problem had been resolved. I now have several new tasks crunching for Einstein. Maybe the new version is fine and something got corrupted here? But this older version is working, so i'll stick with it.
Patrick
Vista 64-bit ultimate, 8GB RAM, Phenom II X4 940, nvidia 8800GT.
RE: ... which was resulting
)
I've used pretty much every version of BOINC in the v6.6 range since v6.6.14, and I think you're misinterpreting how it works.
It does indeed poll every attached project to see if - just perhaps - it might have launched a CUDA app since it last looked: but it also checks for, and downloads, CPU work according to an independent set of priorities.
There may have been reasons why you didn't get CPU work - the changed definition of existing debt values is a case in point - but it wasn't the result of the fishing expedition for potential CUDA work.
RE: There may have been
)
True but it would be nice to turn that off in future Boinc versions, for those of us that chose too of course. Maybe on by default but able to be switched off thru an option. Several people have pointed to it as the cause of their problems, not knowing that it most likely is not. If we could tell them to turn that off and the problem continues, which it will of course, then the trouble shooting could continue. For those that don't use CUDA yet it is a thing they know nothing or very little about.