Re: [Condor-users] slots on SMP machine always in state Owner

[Rok Roškar <roskar@xxxxxxxxxxxxxxxxxxxx>]

A follow up to my previous post about problems with DedicatedScheduler 
on an SMP machine:

here's an excerpt from the StartLog file on the SMP host:

...
8/27 21:34:27 slot5: Request to claim resource refused.
8/27 21:34:27 slot5: Machine requirements not satisfied.
8/27 21:34:27 slot5: State change: claiming protocol failed
8/27 21:34:27 slot5: Changing state: Matched -> Owner
8/27 21:34:27 slot6: Request to claim resource refused.
8/27 21:34:27 slot6: Machine requirements not satisfied.
8/27 21:34:27 slot6: State change: claiming protocol failed
8/27 21:34:27 slot6: Changing state: Matched -> Owner
...

if I submit a regular Vanilla job, the resource is matched and claimed 
just fine - why is it having a problem claiming it when using the 
parallel universe?

Rok

Rok Roškar wrote:
> I've got a 16-cpu SMP machine that is being configured as follows:
> ----------------
> local.config:
> DedicatedScheduler = "DedicatedScheduler@hostname"
> STARTD_EXPRS = $(STARTD_EXPRS), DedicatedScheduler
>
> #SLOTS_CONNECTED_TO_CONSOLE = 16
> #SLOTS_CONNECTED_TO_KEYBOARD = 0
>
> MachineClass = "SMP"
> STARTD_ATTRS = $(STARTD_ATTRS), MachineClass
>
> START = JobClass =!= UNDEFINED && JobClass == "SMP"
>
> Rank = Scheduler =?= $(DedicatedScheduler)
> ------------------------------------------------------------------------------------------
> Even after the machine has been sitting idle for a while (an hour or 
> so), none of the slots are listed as "Unclaimed" instead they're always 
> in state "Owner". Any idea why this is? As a result, jobs submitted to 
> the parallel universe never start because even though the 
> DedicatedScheduler is matched with the correct resources, the job isn't 
> allowed to start (regardless of the START expression ignoring anything 
> to do with keyboard or console activity).
>
> Any thoughts on what I'm missing here are most welcome!
>
> Thanks,
>
> Rok
>
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
>
> The archives can be found at: 
> https://lists.cs.wisc.edu/archive/condor-users/
>