1. (Prepare) Sample PBS Script for running a 1 node - MPI based job:

#!/bin/bash
#PBS -N dns_in
#PBS -l select=1:ncpus=36:vntype=cray_compute
#PBS -l walltime=01:00:00
#PBS -q <Queue Name>
#PBS -j oe
cd $PBS_O_WORKDIR
aprun -n 36 -N 36 ./application

Current Queue Policy :

Queue Name	Number of Nodes
NCMRWF	650
NCMRWF1	350
INCOIS	784
IMD	224
COMPROD	524 (300+ 224 nodes of IMD queue)
serial	1
serial1	1

2. Submit the job to the workload manager(PBS) as :
user@elogin03:~/myjob> qsub myjob.pbs.sh
87493.sdb

3. Checking the status of submitted jobs:
user@elogin03:~/myjob> qstat 87493.sdb

Here are some of the commonly used PBS commands -

Task	Commands
Job submission	qsub [script_file]
Job deletion	qdel [job_id]
Job status (by job)	qstat [job_id]
Job status (by user)	qstat -u [user_name]
Job hold	qhold [job_id]
Job release	qrls [job_id]
Queue list	qstat -Q

Here is the list of some commonly used PBS specific environment variables

Job ID	$PBS_JOBID
Submit Directory	$PBS_O_WORKDIR
Submit Host	$PBS_O_HOST
Node List	$PBS_NODEFILE
Job Array Index	$PBS_ARRAYID

Here is the list of some PBS directive flags -

Queue	-q [queue]
Node Count	-l nodes=[count]
CPU Count	-l ppn=[count] OR -lmppwidth=[PE_count]
Wall Clock Limit	-l walltime=[hh:mm:ss]
Standard Output FIle	-o [file_name]
Standard Error File	-e [file_name]
Combine stdout/err	-j oe (both to stdout) OR -j eo (both to stderr)
Copy Environment	-V
Job Name	-N [name]
Job Restart	-r [y|n]
Resource Sharing	-l naccesspolicy=singlejob
Memory Size	-l mem=[MB]
Tasks Per Node	-l mpiprocs [PEs_per_node]
Job Dependency	-d [job_id]
Job Project	-e [file_name]
Job Arrays	-t [array_spec]
Generic Resources	-l other=[resource_spec]

Presentation Material (13/07/2018)