PBS HPC Plugin

This HPC environment can be selected via --execenv=hpc.pbs. In this HPC environment, SIERRA will run experiments spread across multiple nodes allocated by a scheduler from the PBS family: OpenPBS / PBS Pro, or legacy Torque.

The following table describes the PBS-SIERRA interface. Some PBS environment variables are used by SIERRA to configure experiments during stage {1,2}; if the required ones are not defined SIERRA will throw an error. Because OpenPBS / PBS Pro and Torque expose per-node resources differently, Engine plugins wanting to use the # of allocated cores to set run-time parameters will need to query the environment to see which variable is available:

• PBS_NUM_PPN (Torque),

• NCPUS (OpenPBS / PBS Pro)

The following environmental variables are used in the PBS HPC environment:

PBS-SIERRA interface

Environment variable	SIERRA context
PARALLEL	Used to transfer environment variables into the GNU parallel environment.
PARALLEL_SHELL	Used to set the shell used by GNU parallel to execute all commands in. Overwritten by SIERRA to /bin/bash.
LD_LIBRARY_PATH	Exported by SIERRA via PARALLEL to child GNU parallel processes. Can be undefined when SIERRA starts.
PYTHONPATH	Exported by SIERRA via PARALLEL to child GNU parallel processes. Can be undefined when SIERRA starts.
PATH	Exported by SIERRA via PARALLEL to child GNU parallel processes. Can be undefined when SIERRA starts.
PBS_NODEFILE	Obtaining the list of nodes allocated to a job which SIERRA can direct GNU parallel to use for experiments.
PBS_JOBID	Creating the UUID nodelist file passed to GNU parallel, guaranteeing no collisions (i.e., simultaneous SIERRA invocations sharing allocated nodes) if multiple jobs are started from the same directory.