Are you ready to use Slurm?

Complete this checklist before running the course examples or submitting your own jobs. It will help you avoid common login, storage, software, and job-submission issues on CoSTAR.

For new users, this checklist covers the essential setup steps needed before using the scheduler. For experienced users, it highlights CoSTAR-specific requirements that may differ from other systems.

Confirm your CoSTAR account and network access

Make sure your CoSTAR access has been granted and that you can reach the cluster from your device using an approved access route.

  • Use the recommended access route for your organisation or project.

  • If direct access is not available, use the documented SSH Gateway route for command-line access.

See Log in to the Cluster for the current CoSTAR login methods.

Choose your access method

For new users and external partners, Open OnDemand is the recommended starting point. It provides browser-based file management, shell access, JupyterLab, VS Code, and other interactive applications.

If you are comfortable with the command line, you can also connect by SSH:

ssh <username>@costar-login01.surrey.ac.uk

See Open OnDemand and SSH access for details.

Set up SSH keys

SSH keys are recommended for smoother and more secure access to CoSTAR, the SSH Gateway, and code repositories.

  • Generate a key pair on the machine you use to connect.

  • Deploy the public key to the CoSTAR login node.

  • Keep the private key protected with a passphrase.

See SSH Keys and Deploying SSH Keys.

Set up your code repository

Now is a good time to create or identify the repository you will use for your CoSTAR work. Keep code, job scripts, container recipes, and small configuration files under version control.

Use the repository platform approved for your project. Do not rely on cluster scratch space as the only copy of important code or results.

Get to know Linux

If you’re new to Linux/Unix you may need to take a while to familiarise yourself with how to use the command line interface.

The Ubuntu Command Line for Beginners tutorial is a useful public starting point for learning basic Linux commands and file navigation.

Review storage responsibilities

CoSTAR storage is designed for active computing, not long-term archiving. Before running larger jobs, make sure you understand where your files should live.

  • Keep code and lightweight working files in your home directory.

  • Stage large active datasets and high-throughput job files in scratch or the appropriate project/dataset storage area.

  • Copy important results and final datasets off cluster storage for safe keeping.

Note

Cluster storage is generally not backed up. See Storage Options and Working with files on the cluster.

Check your software approach

CoSTAR users are expected to manage their own software stack where possible. Before submitting production jobs, confirm that your software is available through one of the supported workflows:

  • an Apptainer container, which is the recommended approach for reproducible software environments;

  • a user-managed virtual environment, such as Conda, for lightweight workflows;

  • an Open OnDemand interactive app, where appropriate.

See Software and Requesting New Software.

Prepare your Slurm workflow

Before submitting long-running jobs, check that your job script requests only the resources you need and that it runs successfully on a small test case.

  • Use interactive jobs for debugging and short tests.

  • Use batch jobs for longer, non-interactive workloads.

  • Do not run heavy calculations directly on the login node.

  • Stage large input files before the job starts.

  • Use checkpointing for long model training runs or simulations.

See Job Scheduler, Batch jobs, Interactive jobs, and Checkpointing.

Review the CoSTAR examples

We have prepared a series of examples for you to try.

CoSTAR job submission examples

You should check that you can open the examples and understand the files used by each workflow.

Know where to get support

If you run into problems, use the CoSTAR support channels so the issue can be tracked and handled by the appropriate team.

  • Use the support portal for access, storage, software, or scheduler problems.

  • Include enough detail for support staff to reproduce the issue, such as job IDs, commands, file paths, timestamps, and error messages.

  • Use the CoSTAR Teams channel for community discussion and general questions.

See Available Support.