How to submit job to specific compute-node on the Cray box?


#1

Hi,

It’s using PBS Pro v12.1 on a Cray box with CLE 5.2.

Here are the output of pbsnodes -av

clogin86_12_0
Mom = nid00006,nid00198
ntype = PBS
state = free
pcpus = 16
resources_available.arch = XT
resources_available.host = clogin86_12
resources_available.mem = 33554432kb
resources_available.ncpus = 8
resources_available.PBScrayhost = clogin86
resources_available.PBScraynid = 12
resources_available.PBScrayorder = 86
resources_available.PBScrayseg = 0
resources_available.vnode = clogin86_12_0
resources_available.vntype = cray_compute
resources_assigned.accelerator_memory = 0kb
resources_assigned.mem = 0kb
resources_assigned.naccelerators = 0
resources_assigned.ncpus = 0
resources_assigned.netwins = 0
resources_assigned.vmem = 0kb
resv_enable = True
sharing = force_exclhost
license = l

clogin86_12_1
Mom = nid00006,nid00198
ntype = PBS
state = free
pcpus = 16
resources_available.arch = XT
resources_available.host = clogin86_12
resources_available.mem = 33554432kb
resources_available.ncpus = 8
resources_available.PBScrayhost = clogin86
resources_available.PBScraynid = 12
resources_available.PBScrayorder = 86
resources_available.PBScrayseg = 1
resources_available.vnode = clogin86_12_1
resources_available.vntype = cray_compute
resources_assigned.accelerator_memory = 0kb
resources_assigned.mem = 0kb
resources_assigned.naccelerators = 0
resources_assigned.ncpus = 0
resources_assigned.netwins = 0
resources_assigned.vmem = 0kb
resv_enable = True
sharing = force_exclhost
license = l

I’d like to submit job to certain speicfic compute-node on the Cray box, and tried with these directives, all of them failed.

  1. -l exec_vnode
    #PBS -l select=2:ncpus=16
    #PBS -l vntype=cray_compute
    #PBS -l exec_vnode=(clogin86_12_0:ncpus=8+clogin86_12_1:ncpus=8)+(clogin86_13_0:ncpus=8+clogin86_13_1:ncpus=8)

qsub : read only or insufficient permission Resource_list.exec_vnode

  1. -l host
    #PBS -l select=2:ncpus=16
    #PBS -l vntype=cray_compute
    #PBS -l host=clogin86_12:ncpus=16+clogin86_13:ncpus=16

qsub : Illegal attribute or resource value: clogin86_12:ncpus=16+clogin86_13:ncpus=16

  1. -l nodes
    #PBS -l vntype=cray_compute
    #PBS -l nodes=clogin86_12:ncpus=16+clogin86_13:ncpus=16

qsub : Unknown resource: clogin86_12

Wondering how could I submit job to specifice compute-node on the Cray box.

Thanks for your time

Regards

Jerry


#2

Could you please try this:

  qsub -l select=1:ncpus=1:vntype=cray_compute:host=clogin86_12

Please make sure your chunk request is correct and not to request chunk resources as job wide resources.

Note : chunk are host/vnode level resources, for example output from your pbsnodes -av