Re: [HTCondor-users] Steer job towards high memory nodes

[Vikrant Aggarwal <ervikrant06@xxxxxxxxx>]

Hello Experts,

Any inputs on it.

Â

Thanks & Regards,

Vikrant Aggarwal

On Tue, Sep 17, 2019 at 5:06 PM Vikrant Aggarwal <ervikrant06@xxxxxxxxx> wrote:

Hello Experts,

I want to steer jobs with memory greater than 5342MB to highmemory nodes. HighMemory is added to machine classAD and value is either true or false.Â

_expression_ used to achieve this:Â

NEGOTIATOR_PRE_JOB_RANK = $(machine_rank) \
            + $(disk_speed_rank) \
            - $(cpu_utilization_rank) \
            - $(disk_utilization_rank) \
            + $(memory_rank) \
            - totalcpus

NEGOTIATOR_PRE_JOB_RANK = ifthenelse(requestmemory > 5342 && requestcpus == 1, $(NEGOTIATOR_PRE_JOB_RANK) + 1000000 * (HighMemory =?= true), $(NEGOTIATOR_PRE_JOB_RANK) + (HighMemory =!= true))

But still jobs with high request memory like 7000 MB going to machines with machine classAD of "HighMemory false".Â

It's partitionable slot setup with scheduler splitting enabled. I want to avoid injecting the hard requirement for high memory node in submit file.Â

Can anyone please help to understand what is wrong here.

Thanks & Regards,

Vikrant Aggarwal