Start here · RCC ClusterDocs

RCC Expedition Light: your first 15 minutes

This is the shortest route from a new computer to a safe RCC shell. You do not need to understand the cluster before you begin.

Need an account first? Self-service enrollment is not yet released; until it is, RCC support creates accounts on request (see Request an RCC account). The planned flow is described in Request and activate an RCC account.

Expedition Light is the required first-use path for new users. It deliberately stops after safe access, the basic host and storage model, VS Code, and a small Slurm check. The full RCC Expedition is optional deeper training and introduces containers, Snakemake, and Nextflow later.

RCC is also preparing a browser-first experience for researchers who do not need the command line. There are therefore two starting paths; today only Path A is available.

Path A — command-line access (current)

Use this path now. It gives you SSH, VS Code Remote SSH, direct Slurm commands, workflow development, and automation.

Both checklists use the SSH client already supplied by the operating system. Do not install a separate terminal, Linux virtual machine, or SSH program unless the checklist shows that the built-in client is missing.

After terminal SSH works, follow Use VS Code with RCC for the recommended day-to-day editor, terminal, Git, and remote-file interface.

Path B — browser-first research (not yet released)

Service status — not yet released: RCC Home, Files, and RCC Analysis Notebook/Workflow are not yet available. Use Path A until RCC announces them.

The intended journey is:

RCC Home
   -> Files: upload or choose project data
   -> RCC Analysis: Notebook for interactive exploration
        or RCC Analysis: Workflow for repeatable analysis
   -> Files: inspect/download results

Once released, a browser-only account will be able to work without an SSH public key. RCC web authentication and project membership remain the authority; the browser service submits compute through the governed RCC/Slurm path on your behalf.

What most researchers should remember

1. Your RCC account is your identity. SSH is how you use RCC today; once the browser services are released, browser-only work will not require an SSH key.
2. Files will be the browser data entry/exit surface (not yet released). Durable project inputs and results belong in the project.
3. RCC Analysis Notebook is for exploration. It is planned as Jupyter in a bounded Slurm allocation without manual tunnels or worker selection.
4. RCC Analysis Workflow is for repeatable work. Long, repeated, unattended, or highly parallel analysis belongs in a governed workflow rather than an oversized notebook session.
5. Slurm workers still do the computation. Browser-first changes how you ask for compute, not where compute runs.

The command-line connection model

The command-line connection model is:

Mac or Windows
  -> jump host (automatic forwarding; no working shell)
  -> shell host (prepare and control work)
  -> Slurm allocation (perform computation)

You normally type only ssh shellhost. The ProxyJump line in your SSH configuration takes care of the middle step.

Read the jump-host and shell-host explanation if you need the full command-line mental model.

Where your work belongs

Once released, Files and RCC Analysis should present authorized projects directly, so browser users will not need to type these paths. The paths remain useful reference for developers and reproducibility documentation.

Your primary group records your organizational home. A project is the research collaboration: it has the approved members, data, services, purpose, and lifecycle. Add collaborators to the project; do not move them into another primary group merely to share data.

Turn exploration into reliable analysis

Do not keep an important analysis only in notebook state or shell history. When interactive work becomes repeated, long-running, many-sample, or provenance-critical, turn it into an RCC Analysis Workflow (when released) or use the current script-to-workflow guide.

The resource rule is simple: interactive notebooks should be modest and attended; repeatable/scalable work should become workflows. Requesting more CPU, memory, GPU, or time is not a substitute for measuring what the analysis actually uses.

You are ready when

• terminal SSH reaches the configured RCC target;
• VS Code reaches the same target if you use it;
• you know which project owns the work;
• a small Slurm test completes; and
• repeated analysis is recorded as code/workflow rather than remembered commands.

If you prefer guided, offline training, use RCC Expedition. Returning users from the original IKIM cluster should also read what changed.