Run different tools from hpc using singularity containers and access these from the local browser.
Run the corresponding script as a slurm job from one of the access nodes ie. hpc01 or hpc02.
Scripts are stored in path /usr/share/vt/
sbatch [options] /usr/share/vt/<script>
For example, to launch posit-jupyter with default parameters:
sbatch /usr/share/vt/posit-jupyter
To add different job parameters based on you requirements:
sbatch --partition=<value> --cpus-per-task=<value> --mem=<value> --output=<value> /usr/share/vt/posit-jupyter
In order to use a different image:
sbatch /usr/share/vt/posit-jupyter -i <image_path>
# or
sbatch --export=ALL,IMAGE=<image_path> /usr/share/vt/posit-jupyter
From the job standard output file (default: <scipt_name>.job.<jobid> ), get the instructions:
cat <job_output_file>
To show the standard output file location as well as the job information:
scontrol show job <job_id>
Just follow the fetched instructions to launch the required tool. Followings are the common procedures:
- Do SSH port forwarding from your local computer to and
hpcaccess node: open a local terminal and run the provided command (keep open) - Access the tool from browser with the provided URL (also with credentials if necessary)
- When done using, exit the port forwarding and cancel the job with
scancel -f <job_id>
The scripts offer the following interfaces:
posit-jupyter: Jupyter Lab from Posit
jupyternb: Jupyter Notebook from Posit
vscode: Visual Studio Code
rstudio: Rstudio from the Rocker Project
It is recommended to use the posit-jupyter script as it has more features in one place with the following properties:
- Encrypted https connection
- Multiple versions of R
- Multiple versions of Python
- Option to create R, Python, text or Markdown files