Further PBS Pro environment extension

CERIT-SC continues with the transfer of conventional computing machines (a part of zebra cluster) into the new PBS Pro environment (@wagap-pro). In future, we plan to replace whole current old Torque scheduling system with new PBS Pro.

Machines currently available in a PBS Pro environment are labeled by "Pro" in the PBSMon application https://metavo.metacentrum.cz/pbsmon2/nodes/physical

Frontend zuphux.cerit-sc.cz is set (until at least half of the resources is converted) by default to Torque (@wagap) environment. To activate PBSPro @wagap-pro environment, type the following command after loging on the frontend:

 zuphux$ module add pbspro-client  ... set PBSPro environment

and back 

 zuphux$ module rm pbspro-client   ... return Torque environment

There are no standard resources available in the @arien environment. Although all Torque's queues were disabled next week and it is not possible to submitt new jobs, there are still over 11 thousand of jobs that can not be computed at @arien.

We started migration of jobs with compatible setting to CERIT-SC Torque (@wagap) environment. But unfortunatelly, jobs with special setting or property not available in the CERIT-SC Torque environment (GPU, location outside Brno, array, etc.), can not be migrated automatically, they need to be rewrited for PBS Pro (@arien-pro) and resubmited by the job owner.

All the frontends (except wagap) are set to PBS Pro environment @arien-pro by default.

Note: Main difference of the PBS Pro:

New select syntax:

qsub -q uv -l select=1:ncpus=48:mem=20gb:scratch_local=20gb -l walltime=1:00:00 skript.sh

Always specify
- walltime in [hh:mm:ss] format
- size and type of the scratch <scratch_local|scratch_ssd|scratch_shared>

Documentation for new scheduling system PBS Professional can be found at wiki https://wiki.metacentrum.cz/wiki/PBS_Professional

MetaCentrum & CERIT-SC

Ivana Křenková, Tue Mar 28 21:39:00 CEST 2017