Re: [HTCondor-users] Rank with Partitionable and Dynamic Slots

[Nathan Yehle <nyehle@xxxxxxxxxxxxxxxxxxxxx>]

Hi,
Rank works, we use rank plus these three settings you might want to try, 
perhapse your preempt and preemption_requirements expressions are never 
matching, the defaults didnt work with rank for me.  If 
MAXJOBRETIREMENTTIME is high it can take a long time for the lesser 
priority job to retire.

 #rank to favor 120G jobs
RANK = $(RANK) + ( RequestMemory >= 120000)
PREEMPTION_REQUIREMENTS = MY.TotalJobRunTime > MAX_PREEMPT && ( 
RequestMemory < 120000)|| MY.TotalJobRunTime > ( 240 * (60 * 60) ) ( 
RequestMemory >= 120000)
MAXJOBRETIREMENTTIME = 0
PREEMPT = CpuBusyTime > 10 * $(MINUTE)

-Nate

On Mon, 25 Aug 2014, matthias.roehm@xxxxxxxxxx wrote:

> Hello,
>
> I just started to play around with Dynamic Provisioning in Condor 8 as 
> described in section 3.5.10.7 of the manual.
> Submitting and executing jobs works fine.
> Only the RANK expression within the submit file is completely ignored.
> As an example: Submitting a job with "rank = TotalCpus" will not result in 
> executing on the machine with the most CPU-Cores, instead the job will run on 
> the machine with the least CPU-Cores.
>
> When I remove dynamic slots from the config file, RANK is working fine again.
> I see this behavior with both Condor 8.2.1 and 8.3 running on Scientific 
> Linux 6.4.
>
> Is RANK supposed to work with dynamic slots?
>
> Best,
> Matthias
>
> Here is my conf file for the compute servers:
> ######################################################################
> use feature : GPUs
> GPU_DISCOVERY_EXTRA = -extra
>
> NUM_SLOTS=1
> NUM_SLOTS_TYPE_1=1
> SLOT_TYPE_1=100%
> SLOT_TYPE_1_PARTITIONABLE=true
>
> DAEMON_LIST = MASTER STARTD
> ######################################################################
>
> And here a test sub file:
> ######################################################################
> universe = vanilla
> executable = run.sh
> rank = TotalCpus
> request_cpus = 4
> request_GPUs = 1
> log = test$(Cluster).log
> output = test$(Cluster).out
> error = test$(Cluster).err
> queue
> ######################################################################
>
>
>
> _______________________________________________
> HTCondor-users mailing list
> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/htcondor-users/