Skip to content
Spike Documentation

Terminal access

The simplest way to look around inside a running workspace is to open a Bash shell with the Run:ai CLI. This connects directly to the workspace’s main pod; you do not need to install or configure an SSH server in the container.

Before connecting, install and sign in to the Run:ai CLI, connect eduVPN when required, and confirm in the Run:ai portal that the workspace is Running.

Open a shell in a running workspace

Section titled “Open a shell in a running workspace”

Find the workspace name, then open a shell:

Terminal window
runai workspace list --project PROJECT_NAME
runai workspace bash WORKSPACE_NAME --project PROJECT_NAME

You are now running Bash inside the workspace’s main pod. From here you can inspect the filesystem, check the mounted data, or verify that the GPU is visible:

Terminal window
pwd
ls -la
nvidia-smi

Type exit to leave the shell. This disconnects your terminal but does not stop the workspace. If the workspace has more than one pod, select one explicitly:

Terminal window
runai workspace bash WORKSPACE_NAME \
  --project PROJECT_NAME \
  --pod POD_NAME

See NVIDIA’s runai workspace bash reference for the available options.

Advanced: Connect to the SSH container

Section titled “Advanced: Connect to the SSH container”

SSH is only available when the workspace uses an image that runs an SSH server. Spike provides an SSH-container example for this purpose; a regular workspace image does not need SSH. The Podman and Harbor guide links all files needed to build the example.

Once that SSH-enabled workspace is Running:

  1. Connect eduVPN when required by your location, then run runai login.

  2. Keep the port forward open in one terminal:

    Terminal window
    runai workspace port-forward WORKSPACE_NAME \
      --project PROJECT_NAME \
      --port 2222:2222

  3. In a second terminal, connect with the private key matching the public key added when the SSH-container image was built:

    Terminal window
    ssh -i ~/.ssh/id_ed25519 developer@localhost -p 2222

Keep the forwarding terminal open while using SSH. Reconnect by checking that the workspace is still Running, then start the port forward again.

Use VS Code Remote - SSH

Section titled “Use VS Code Remote - SSH”

Once the command-line SSH test succeeds, the same forwarded connection can be used by Visual Studio Code Remote - SSH. Install the Remote - SSH extension locally and add a host entry to ~/.ssh/config:

Host spike-workspace
  HostName localhost
  User developer
  Port 2222
  IdentityFile ~/.ssh/id_ed25519
  IdentitiesOnly yes

Keep the Run:ai port-forward running, open the VS Code Command Palette, choose Remote-SSH: Connect to Host…, and select spike-workspace. Open the project directory on the PVC rather than keeping the only copy of code in the container’s temporary filesystem.

VS Code installs a small server component in the remote environment and may need outbound HTTPS during first use. A minimal or non-glibc image may not meet its prerequisites. If setup fails, use runai workspace bash while you and your SCC AI Engineer review the image; do not add a password, root login, or an unreviewed SSH daemon configuration as a workaround.

For moving files rather than opening a shell, follow the complete File transfer guide.