Seite 4 von 93

Re: Long running work unit

Verfasst: 14.05.2010 02:42
von Conan
Conan hat geschrieben:31 hours 28%
31.5 hours 4%
41.5 hours 17%

Work unit appears to have reset itself, the estimated run time has changed from 110 hours to 244 hours.

I will post the link when it finishes.

(Computer is an AMD Phenom II X4 955 @3.2 GHz, running Windows XP)
Now up to 56.12 hours and 43% (which gives 130 hours run time)

This is the Work Unit due on the 22/5/10 so I should easily make it.

It has an project estimated run time of 1d 19h 26m.

I hope it hurries up and finishes as another 12 work units have now downloaded and I need to start on them.

Edit:--- Has jumped from about 50% to 100% and finished after 58 hours.

Re: Long running work unit

Verfasst: 15.05.2010 11:38
von Roland Schneider
Now I have one WU, which aborted because the maximum time elapsed exceded.
This WU aborted on my maschine (ID 444) with aproximately 20 percent done.

Re: Long running work unit

Verfasst: 15.05.2010 14:28
von yoyo
I increased the max cpu time for all unset workunits by factor 10.
yoyo

Re: Long running work unit

Verfasst: 15.05.2010 19:17
von Ananas
Oops ... bad config on this one : http://www.rnaworld.de/rnaworld/workuni ... id=1020526

Code: Alles auswählen

    <rsc_fpops_est>1052073389927580.000000</rsc_fpops_est>
  <rsc_fpops_bound>1000000000000000.000000</rsc_fpops_bound>
Estimated is more than the limit.

BOINC should really reject something like this on server side already.

Re: Long running work unit

Verfasst: 16.05.2010 17:20
von Michael H.W. Weber
Please note that RNase_MRP and the rRNA WUs are really the longest we currently have to offer.

Michael.

Re: Long running work unit

Verfasst: 16.05.2010 23:35
von Ananas
Michael H.W. Weber hat geschrieben:Please note that RNase_MRP and the rRNA WUs are really the longest we currently have to offer.

Michael.
Problem is not the runtime itself, it's more the estimated runtime in relation to the maximum runtime. The allowed maximum should at least be twice (preferred 10x) as long as estimated, in our case (different SSE instruction sets supported) even a bit more.

BOINC should issue a warning when the WU is submitted with inconsistant runtime values and even reject it when the estimate is higher than the maximum.

Re: Long running work unit

Verfasst: 17.05.2010 13:34
von Michael H.W. Weber
Ananas hat geschrieben:BOINC should issue a warning when the WU is submitted with inconsistant runtime values and even reject it when the estimate is higher than the maximum.
Agreed. But this must be addressed at David Anderson and his team, I guess. 8)

Michael.

Re: Long running work unit

Verfasst: 22.09.2010 03:29
von darwincollins
I created a thread about this a few weeks back, and it was suppoused to be rare that this happens. Well, I just aborted another 3 WUs.

Anyone closer to a fix this?

Re: Long running work unit

Verfasst: 22.09.2010 15:01
von Michael H.W. Weber
Could you please specify what exactly you are talking about?

Michael.

Re: Long running work unit

Verfasst: 24.09.2010 04:52
von darwincollins
cmsearch 0.14
cms_GA[e30-50MB_Lin64f]_Equus-caballus-(horse)_CM000389.lin.EML_RF00028_Intron_gpI_1284623755_2217_5

Elapsed 22.06:36
To Completion: 92:06:44

(its using 1 core of dual 3ghz quad core server)

Is this another candidate for abort, or should I just let it keep cranking away?

Re: Long running work unit

Verfasst: 24.09.2010 06:40
von MReed
Since it is an Intron WU, keep it running as long as it consumes CPU time. If it stops using the core AND does not show "Waiting for Memory" you should abort it.

Re: Long running work unit

Verfasst: 26.09.2010 08:04
von darwincollins
They were still running. One did finish, it said that it took 15min.

Does anyone have any clues if we are talking about bugs in the code, or just simply work that will take alot of time ?