Re: [HTCondor-users] Idle jobs exist while resources are available

[Greg Thain <gthain@xxxxxxxxxxx>]

On 10/03/2016 10:05 AM, Zhuo Zhang wrote:

Hi,

We have a small test cluster with 5 machines. After submitting jobs to the test cluster, I always see 5 slots unclaimed even there are jobs waiting in the queue. I am not an administrator but a user, just suspect there are something wrong with our condor configurations. See the experiments I did below, can someone explain why there are jobs waiting while the resources are still available? Is the behavior expected?

3. But condor_status showed there are still 5 slots unclaimed.

rhw1193:25} condor_status
Name               OpSys      Arch   State     Activity LoadAv Mem     ActvtyTime

slot1@xxxxxxxxxxxx LINUX      X86_64 Unclaimed Idle      0.140 146991  4+19:01:06
slot1@xxxxxxxxxxxx LINUX      X86_64 Unclaimed Idle      0.020 146991 32+22:31:04
slot1@xxxxxxxxxxxx LINUX      X86_64 Unclaimed Idle      0.060 146990 32+22:22:23
slot1@xxxxxxxxxxxx LINUX      X86_64 Unclaimed Idle      0.830 146991  4+18:58:47
slot1@xxxxxxxxxxxx LINUX      X86_64 Unclaimed Idle      0.300 136879 18+17:51:51

If these five slots are the only Unclaimed ones, then your configuration is probably OK.  These slots are probably partitionable slots, which means that they hold the "leftover" resources (cpu, memory, disk, etc.) that any running dynamic slots under them (the slot1_xx slots) are using.  If you run

condor_status -af Name Cpus

condor will tell you how many cpus are in each slot, and I bet that the partitionable slots have exhausted their cpus, and there are 0 there.

-greg