Manage sessions
After creating a Workbench session, you can pause it to release compute while keeping your storage, resume it to pick up where you left off, edit its configuration, or delete it entirely.
Prerequisites
- An existing Workbench session (see Create a session)
Open a session
When a session is in Running status, click the Open button on the session card or detail page. Your notebook environment (JupyterLab, VS Code, or RStudio) opens in a new browser tab.
Pause a session
Pausing releases compute resources (CPU, memory, GPU) while retaining all storage volumes. Your files, installed packages, and workspace state are preserved.
- Click the ... menu on the session row or open the session detail page.
- Click Pause.
- The session transitions from Running to Paused.
Pause sessions you are not actively using to reduce costs, especially when GPU resources are allocated.
Resume a session
- Click the ... menu on a paused session.
- Click Resume.
- The session transitions through Pending (re-provisioning compute, re-mounting storage) back to Running.
Your workspace is exactly as you left it.
Edit a session
You can change the CPU, memory, GPU allocation, environment variables, and mounted volumes on an existing session.
- Click the ... menu on the session row and select Edit, or click the Edit button on the detail page.
- Modify the fields you need to change.
- Click Save.
Some changes (like compute size) require the session to be paused first. The UI will prompt you if a pause is needed.
Delete a session
Deleting a session tears down both compute and storage volumes. This is irreversible.
- Click the ... menu and select Delete.
- Confirm the deletion in the dialog.
If you might need the session again, prefer Pause over Delete. Paused sessions retain storage at minimal cost.
Session lifecycle states
| State | Meaning |
|---|---|
| Pending | Provisioning compute, pulling images, mounting storage |
| Running | Healthy and accepting work; Open button available |
| Paused | Compute released, storage retained |
| Failed | Last reconcile errored; check the Activity log |
| Terminating | Being deleted |
| Unknown | State could not be determined |
Monitor costs
The session detail page shows a cost line with hourly rate, accumulated spend, and GPU utilization. An "idle" badge appears if GPU utilization drops below 5% for more than one hour.
Related
- Create a session
- Session lifecycle for detailed state transition diagrams