Re: [Condor-users] slot1: Job Requirements check failed!

[Matthew Farrellee <matt@xxxxxxxxxx>]

On 12/10/2010 02:23 AM, Carsten Aulbert wrote:
> Hi
>
> this is a follow-up to my last email, but addressing a completely different
> issue.
>
> I'm using a job submit file as follows:
> Executable     = sleep.gpu.sh
> Arguments      = 10 $$(GPU_DEV) $$(GPU_NAME) $$(GPU_CAPABILITY) \
> $$(GPU_GLOBALMEM_MB) $$(GPU_MULTIPROC) $$(GPU_NUMCORES) $$(GPU_CLOCK_GHZ) \
> $$(GPU_CUDA_DRV) $$(GPU_CUDA_RUN)
> Error   = logs/err.$(Process)
> Output  = logs/out.$(Process)
> Log = /local/user/carsten/foo.log
> Requirements = GPU_CAPABILITY>= 1.9
> +WantGPU=True
> Universe = vanilla
> Queue 1
>
> where sleep.gpu.sh is only printing out the arguments and sleeping for $1
> seconds.
>
> With "Requirements = GPU_CAPABILITY>= 2.0" I'm trying to steer it to a
> machine which has this one set. It kind of works, the match is made, but when
> the startd wants to start the job, it just says "slot1: Job Requirements check
> failed!" and goes back to idle (full debug startLog attached).
>
> $ gpu010:/var/log/condor# grep -i require /tmp/StartLog
> AutoClusterAttrs =
> "JobUniverse,LastCheckpointPlatform,NumCkpts,Scheduler,Owner,NeedGpu,WantGPU,DiskUsage,ImageSize,RequestMemory,FileSystemDomain,Requirements,NiceUser,ConcurrencyLimits"
> Requirements = (GPU_CAPABILITY>= 2.000000)&&  (Arch == "X86_64")&&  (OpSys ==
> "LINUX")&&  (Disk>= DiskUsage)&&  ((Memory * 1024)>= ImageSize)&&
> ((RequestMemory * 1024)>= ImageSize)&&  (TARGET.FileSystemDomain ==
> MY.FileSystemDomain)
> Requirements = (START)&&  (IsValidCheckpointPlatform)
> AutoClusterAttrs =
> "JobUniverse,LastCheckpointPlatform,NumCkpts,Scheduler,Owner,NeedGpu,WantGPU,DiskUsage,ImageSize,RequestMemory,FileSystemDomain,Requirements,NiceUser,ConcurrencyLimits"
> Requirements = (GPU_CAPABILITY>= 2.000000)&&  (Arch == "X86_64")&&  (OpSys ==
> "LINUX")&&  (Disk>= DiskUsage)&&  ((Memory * 1024)>= ImageSize)&&
> ((RequestMemory * 1024)>= ImageSize)&&  (TARGET.FileSystemDomain ==
> MY.FileSystemDomain)
> Requirements = (START)&&  (IsValidCheckpointPlatform)
> 12/10 10:43:42 slot1: Job Requirements check failed!
>
>
> I'm not quite sure what is causing Condor to not start the job, at first I
> thought it might be the floating-point comparison, but even with
>
> Requirements = GPU_CAPABILITY>= 1.9
>
> it matches, but does not start.
>
> Any ideas?
>
> Cheers
>
> Carsten

You might try wrapping your Requirements in a debug(),

 $ condor_qedit 1702098.0 Requirements 'debug((GPU_CAPABILITY >= 
2.000000) && (Arch == "X86_64") && (OpSys == "LINUX") && (Disk >= 
DiskUsage) && ((Memory * 1024) >= ImageSize) && ((RequestMemory * 1024) 
>= ImageSize) && (TARGET.FileSystemDomain == MY.FileSystemDomain)
)'

Don't do it too much though, you'll spam the Sched&Negotiator&StartLogs

Best,


matt