7 Workflows and Best Practices
This chapter contains information about how we run analysis locally and on UCSB’s HPC clusters.
7.1 Commands on Pod
| Command | Use |
|---|---|
ssh {username}@pod-login1.cnsi.ucsb.edu |
login to pod |
srun --pty bash -i |
run in a tmux session, launch an interactive job on a normal memory node (add to the queue) that doesn’t respect SBATCH commands in the sh files run |
sbatch script.sh |
launch a non-interactive job on a normal memory node (add to the queue) |
sbatch -p largemem script.sh |
launch a non-interactive job on a high memory node (add to the queue) |
squeue -u {username} |
print information regarding requested jobs such as JOBID, NAME, NODES, etc. |
ssh node48 |
open terminal in the appropriate node number such as node 48 |
scontrol show job {jobID} |
show detailed info about your job, such as endtime |
top -u {username} |
show node memory usage and only your processes, with %MEM column showing each process’s memory usage (htop is not available on Pod) |
module load R/4.1.3 gdal/2.2.3 proj/5.2 |
load modules in one of the terminals on the job node |
R |
open R module in terminal |
q() |
exit out of R module, switch wd to ~ |
sinfo -o "%n %m" |
shows the total amount of memory per node for all nodes on the server |
sinfo -o "%n %m %C" | awk '$2 >= 512000' |
shows the core availability of each node that has at least 500GB |