TAMU Supercomputing Facility:HPRC:Batch Translation
Batch Job Translation Guide
Contents
Overview
Over the last couple of years, TAMU HPRC has had a few different job managers. Our clusters in the recent years are listed below along with their job manager.
- Ada: LSF
- Terra: Slurm
- Grace: Slurm
- Curie (RETIRED): LSF
- Eos (RETIRED): PBS/Torque
- Neumann (RETIRED): Load Leveler
This page contains helpful translation tables for the purposes of moving jobs from one cluster to another. The job managers in the tables below are LSF, Slurm, and PBS/Torque, the ones used on our three main clusters: Ada, Terra, and Grace.
For more information on our clusters, please see their respective pages:
Translation Tables
User Commands
The table below contains commands that would be run from the login nodes of the cluster for job control and monitoring purposes.
| User Commands | LSF | Slurm | PBS/Torque |
|---|---|---|---|
| Job submission | bsub [script_file] | sbatch [script_file] | qsub [script_file] |
| Job deletion | bkill [job_id] | scancel [job_id] | qdel [job_id] |
| Job status (by job) | bjobs [job_id] | squeue --job [job_id] | qstat [job_id] |
| Job status (by user) | bjobs -u [user_name] | squeue -u [user_name] | qstat -u [user_name] |
| Queue list | bqueues | squeue | qstat -Q |
Environment Variables
The table below lists environment variables that can be useful inside job files.
| Environment Variables | LSF | Slurm | PBS/Torque |
|---|---|---|---|
| Job ID | $LSB_JOBID | $SLURM_JOBID | $PBS_JOBID |
| Submit Directory | $LSB_SUBCWD | SLURM_SUBMIT_DIR | $PBS_O_WORKDIR |
Job Specifications
The table below lists various directives that set characteristics and resources requirements for jobs.
| Job Specification | LSF | Slurm | PBS/Torque |
|---|---|---|---|
| Script directive | #BSUB | #SBATCH | #PBS |
| Node Count | N/A ( Calculated from: CPUs/CPUs_per_node ) |
-N [min[-max]] | -l nodes=[count] |
| CPUs Per Node | -R "span[ptile=count]" | --ntasks-per-node=[count] | -l ppn=[count] |
| CPU Count | -n [count] | -n [count] | N/A ( Calculated from: CPUs_per_node * Nodes ) |
| Wall Clock Limit | -W [hh:mm] | -t [min] or -t[days-hh:mm:ss] | -l walltime=[hh:mm:ss] |
| Memory Per Core | -M [mm] AND -R "rusage[mem=mm]" * Where mm is in MB |
--mem-per-cpu=mem[M/G/T] | -l mem=[mm] * Where mm is in MB |
| Standard Output File | -o [file_name] | -o [file_name] | -o [file_name] |
| Standard Error File | -e [file_name] | -e [file_name] | -e [file_name] |
| Combine stdout/err | (use -o without -e) | (use -o without -e) | -j oe (both to stdout) OR -j eo (both to stderr) |
| Event Notification | -B and/or -N * For Begin and/or End |
--mail-type=[ALL/END] * See 'man sbatch' for all types |
-m [a/b/e] * For Abort, Begin, and/or End |
| Email Address | -u [address] | --mail-user=[address] | -M [address] |
| Job Name | -J [name] | --job-name=[name] | -N [name] |
| Account to charge | -P [account] | -account=[account] | -l billto=[account] |
| Queue | -q [queue] | -q [queue] | -p [queue] |
Advanced Documentation
This guide only covers the most commonly used options and useful commands.
For more information on LSF batch options: Official LSF V9.1.2 Documentation
Or see easier-to-read LSF docs: LSF V9.1.2 Documentation
For more information on Slurm batch options: Official Slurm Documentation
