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

Re: [Condor-users] Job stays idle when using Dynamic Slots



Colak Birol wrote:
> Hi,
> 
> I am using Condor 7.2.2 on RHEL3. I have computing nodes with 2 dualcore 
> CPUs, so I configured one dynamic slot with 4 CPUs on each node:
> 
> NUM_SLOTS = 1
> SLOT_TYPE_1 = cpus=4
> NUM_SLOTS_TYPE_1 = 1
> SLOT_TYPE_1_PARTITIONABLE = true
> 
> When I submit jobs over the SOAP interface, adding following ClassAds to 
> the Job, it runs fine (the job gets running after some minutes on a 
> dynamically created slot).
> 
> RequestCpus = 1
> RequestMemory = ceiling(ImageSize / 1024.000000)
> RequestDisk = DiskUsage
> 
> But when I say RequestCpus = 2, the job stays idle forever. I assume it 
> has something to do with the AutoClusterAttrs, which is
> 
> AutoClusterAttrs = 
> "JobUniverse,LastCheckpointPlatform,NumCkpts,RequestCpus,RequestMemory,RequestDisk,Requirements,NiceUser,ConcurrencyLimits"
> 
> Anyone has similar problems or a solution?
> 
> Best Regards,
> Birol

The AutoClusterAttrs shouldn't make a difference.

Does condor_q -better-analyze tell you anything useful?

To get a deeper understanding of why the job is idle you can look at the
StartLog and SchedLog, both should reference the job id.

Best,


matt