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

Re: [Condor-users] X86_64 Cluster refuses to run INTEL jobs?



On Fri, Jun 24, 2005 at 09:05:30AM -0400, Eric Warnke wrote:
> Robert,
> 
> Yes you are correct I am using a globus toolkit and no I don't think 
> this is isolated to globus since I have not used it to submit jobs.  If 
> I take a simple job submission that works on the X86_64 headnode and run 
> it through condor_submit on the i386 node it fails to match to any 
> host.  If I take the same job and change the requirements to 
> ((Arch=="X86_64")||(Arch=="INTEL")) it will run.  I have found the 
> section of condor.pm that I would need to modify, but that seems like an 
> ungly hack to me since once globus supports the distinction of i386 and 
> x86_64 it will more than likley run into trouble.
> 
> My problem is that the negotiator SHOULD be matching INTEL jobs with 
> X86_64 execution nodes.
> 

In jobs you submit, you'll have to include a
((Arch=="X86_64")||(Arch=="INTEL"))

expression if you want it to match both, or change the configuration 
files on your X86_64 machines to include

ARCH = "INTEL"

The default is going to remain X86_64 machines report X86_64, and IA32
machines report "INTEL" for now - there are people who don't want
IA32 jobs to run on X86_64, so it's safer for us to not run them
by default. (ie we'd rather have jobs not run than run with errors)

-Erik