[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] LoadAvg vs. CondorLoadAvg



On Fri, Apr 17, 2015 at 07:45:25AM -0700, Evan Sarmiento wrote:
> Hi,
> 
> 
> Iâm trying to figure out how this is possible. When I run:
> 
> 
> condor_status -format "%d " '(LoadAvg - CondorLoadAvg)â
> 
> 
> 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Add a '\n' in the format if you want separate lines (pipe through "sort | uniq -c"
to count different results).

> In a job, when I run a CPU intensive job, like, âddâ, or âmd5sumâ, I see the LoadAvg increasing, but CondorLoadAvg always equals âLoadAvg.â This makes the CpuBusy and CpuBusyTime always equal âfalseâ

CpuBusy is meant to detect "foreign" (to Condor) activity, for example if
a machine is also used as a desktop computer or serving other tasks -
Condor can then decide not to schedule jobs to it, or even preempt
running (Condor) jobs.
So I think this is intentional and not a bug - Condor uses slots for
bookkeeping of resources and won't oversubscribe them.

- S

-- 
Steffen Grunewald * Cluster Admin * steffen.grunewald(*)aei.mpg.de
MPI f. Gravitationsphysik (AEI) * Am MÃhlenberg 1, D-14476 Potsdam
http://www.aei.mpg.de/ * ------- * +49-331-567-{fon:7274,fax:7298}