Quick Reference (Cheat Sheet)

Note

This page is a quick reference guide to be used once you are familiar the HEC User Guide. Many entries contain a link to more detailed information on these commands

Login, filestore and data transfer

ssh USERNAME@wayland.hec.lancs.ac.uk	login from a linux desktop
gpfsquota	show usage for for HEC filestores
scp -r /path/to/my/directory USERNAME@wayland.hec.lancs.ac.uk	upload a directory to the HEC from a linux desktop
scp -r /path/to/my/directory USERNAME@wayland.hec.lancs.ac.uk	upload a directory to the HEC from a linux desktop
smbclient -k -D folder-path //luna/faculty-code	connect to a research folder on LUNA. Use put and get to upload/download

Job Submission

sbatch JOBSCRIPT	submit file JOBSCRIPT as a batch job
srun --pty bash -i	submit an interactive job
scancel JOBID	cancel job number JOBID

Job Monitoring

squeue --me	shows all running and pending jobs
qcgtop -u USERNAME	shows CPU and memeory usage summaries for all running jobs
qtop -u USERNAME	shows top-style information on all running jobs
qquota	shows job resource quotas (CPUs, memory and GPUs)
qacct -j JOBID	shows summary of completed job JOBID
qslots	shows available CPU resource on the cluster
qgpuslots	shows available GPU resource on the cluster

Enviroment Modules

module avail	show all modules
module av PATTERN	show available modules matching PATTERN
module add MODULENAME	add MODULENAME to the current shell environment
module rm MODULENAME	remove MODULENAME from the current shell environment
module list	show all modules loaded into the current shell environment
module purge	remove all modules from the current environment

Job Type Templates

#SBATCH -p serial #SBATCH --mem=SIZE	a serial job requesting SIZE memory (eg. 1G = 1 gigabyte)
#SBATCH -p parallel #SBATCH --cpus-per-task=N #SBATCH --mem=SIZE	a single node parallel job with N cores on the same compute node, and SIZE memory
#SBATCH -p parallel #SBATCH --nodes=N #SBATCH --exclusive #SBATCH -C node_type=X	a multi node parallel job requesting N nodes of compute node type X
#SBATCH -p gpu-short #SBATCH --gres=gpu:1 #SBATCH --mem=SIZE #SBATCH --cpus-per-task=N	submit a gpu job to the gpu-short queue, requesting N cores, SIZE memory and a single GPU
#SBATCH --mail-type=END,FAIL #SBATCH --mail-user=youraddress@lancaster.ac.uk	send an email notification when a job finishes or fails
#SBATCH -a x-y:z	tuirn the job into a job array running from task ID x through to y in steps of z