Auto-Stop on Idle

Auto-stop automatically shuts down a container after its CPU has been idle for a timeout you choose. It’s the easiest way to avoid paying for compute you aren’t using — leave a notebook or dev environment running and it will power itself off when you step away.

What it does

When auto-stop is enabled on a container, Carolina Cloud polls its CPU every 5 minutes. If CPU usage stays below 5% for the entire timeout window, the container is stopped cleanly — the same way it would be if you clicked Stop yourself. Your disk and data are preserved; only compute billing stops.

Supported instances

Auto-stop is available for containers only — every flavor (base, genomics, datascience, marimo, datasciencemarimo, rstudioserver, rgeospatial). It is not available for VMs.

Enabling auto-stop

On your instance card in the dashboard:

1. Flip the card (the circular arrow in the top-right corner) to reveal the back.
2. Check Auto-stop when idle.
3. Click the pencil next to the timeout to change it (default is 30 minutes).

You can also set auto-stop via the API:

curl -X PATCH https://console.carolinacloud.io/api/instance/<uuid>/ \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"auto_stop_enabled": true, "auto_stop_minutes": 60}'

When auto-stop is enabled, a small clock chip appears on the front of the instance card showing the timeout.

How idle is measured

• Threshold: CPU usage under 5% counts as idle.
• Check interval: Carolina Cloud polls every 5 minutes.
• Timeout: 1–1440 minutes (up to 24 hours). Default 30.
• Grace window: The idle clock is reset whenever the container starts, restarts, or auto-stop is first enabled — so a freshly-enabled toggle always gets a full timeout before anything happens.
• Unreachable host: If the container’s host can’t be polled, the event is ignored rather than counted as idle.

If CPU crosses back above 5% at any check, the idle clock resets to zero.

Restarting an auto-stopped container

Auto-stopped containers show the EXITED state, identical to a manual stop. Start them again from the dashboard, the CLI (ccloud start <name>), or the API. All data is preserved on the container’s NVMe disk.

Billing impact

• Running: Full compute + storage cost.
• Auto-stopped: Storage only — compute billing halts the moment the container is stopped, metered to the partial interval.

See Pricing & Billing for full rate details.

When to use it

• Notebook servers (Marimo, Data Science, RStudio) that you open and close throughout the day.
• Dev containers you SSH into for a few hours at a time.
• Scheduled jobs where you restart the container on demand and it should shut itself off afterward.

For always-on workloads (a long-running training job that dips to low CPU between epochs, a service that must stay reachable), leave auto-stop disabled or set the timeout long enough to cover normal quiet periods.