No work??
1 Feb 2009 15:50:42 UTC
Topic 194176
(moderation:
Over the last three days, have not received one job from Einstein whilst it has been arriving from others (ABC and CPDN)
Am running BOINC 6.2.12 in Ubuntu 64 bit and am wondering if it is just me (likely) or something within this version of BOINC.
Am permanently connected to the net, on a dual boot (only double up on CPDN) Any suggestions/help greatly appreciated since Seti is on "holidays"
No work??
)
Well, whinge and thou shalt receive.
Normal service has been resumed.
BOINC works on a Debt system.
)
BOINC works on a Debt system. If one project has had more time that it should, it will pause for a while. This is very normal operation. You may see this with any project. CPDN often will wait several days after finishing a project before getting a new one (some people have observed up to a couple of weeks).
Nothing to be alarmed about. Micromanaging causes issues, let the software do what it needs.
Haven't gotten any new
)
Haven't gotten any new Einstein work for days. I'm running BOINC 6.6.9:
Processor: 2 GenuineIntel Intel(R) Core(TM)2 Duo CPU T9400 @ 2.53GHz [x86 Family 6 Model 23 Stepping 6]
Processor features: fpu tsc pae nx sse sse2 pni mmx
OS: Microsoft Windows Vista: Ultimate x86 Editon, Service Pack 1, (06.00.6001.00)
Memory: 3.00 GB physical, 6.19 GB virtual
Einstein shares 1/3 of my resources with SETI and Climateprediction getting the other 2/3. SETI is actually using my CUDA, so has little impact on the real CPU. I run BOINC all day, so I don't understand why Einstein gives me no work anymore. My system can usually finish one or more Einstein WUs/day.
RE: ... I don't understand
)
Last weekend there was a fairly protracted problem with Seti. During that time Einstein probably had a field day and probably downloaded a lot more tasks than its normal entitlement. This would seem to be borne out by the fact that your current E@H RAC is much larger than your current Seti RAC.
I would imagine that BOINC is simply favouring Seti at the moment now that Seti tasks are freely available again. Once the balance has been restored, E@H will be allowed to download new tasks once more. This is likely to be normal BOINC behaviour and therefore nothing to worry about.
Did you run out of Seti tasks at any stage during the Seti outage? Have you checked the relative levels of LTD (long term debt) for all projects in your state file (client_state.xml) to see if debt imbalance is indeed the cause of your situation?
Cheers,
Gary.
RE: Did you run out of Seti
)
If only it was that easy these days, but with 6.6 there is a whole new debt system, new client side CPU scheduler and work fetch module, it's not what it used to be.
There's still LTD and STD, but also CPU_backoff_time and CUDA_backoff_time to consider. All debts are now under zero, with zero being the end point. There's no debt above it anymore. ;-)
It's still that the project with the least amount of LTD will fetch work, but even that's not completely working in 6.6.9 ... it's better in 6.6.11, hence why people who are using these development versions should update to the next one when it's out, or not use it at all if they do not plan to be testing them.
That said, the present recommended 6.4.6 is a little broken. Hopefully Rom will fix that later today once he's awake.
RE: If only it was that
)
Thanks very much for noticing he was using an alpha test version of BOINC and thanks very much for filling in some of the details as to why I won't be using it any time soon :-).
I guess the moral of the story is that people having BOINC related problems should ask their questions on the BOINC boards where much more knowledgeable people can give them a much better answer. I've certainly got no knowledge of what's happening with major changes like this in the development versions. I guess I'll find out one of these days if E@H ever gets around to testing a CUDA app - maybe in 2010 if I'm lucky :-).
Cheers,
Gary.
RE: I guess I'll find out
)
One thing Jim Potter didn't mention, but could easily check, is the BOINC message log in the Advanced view of BOINC Manager.
Look back for the last time there's a log entry concerning Einstein contacting the servers. If it says something like "Not requesting work" or "Requesting 0 seconds", then the guesses about debt payback are probably right.
But if the messages suggest that BOINC requested work, but failed to receive it, then it could be an Einstein problem and it's worth posting again about it - preferably with a copy of the messages in question.