Skip to content

Start and stop your server safely

Distinguish graceful shutdown from force stop, and wait for confirmed state.

Audience: server users with control.power. Controls also depend on server state and current operations.

Choose an action

Action Meaning Before you use it
Start Request startup of a stopped machine. Check installation and host availability.
Shutdown Ask the guest to stop gracefully. Save work and allow applications to finish.
Restart Request a guest restart. Warn active users and save work.
Kill Force the machine off. Use only after assessing the data-loss risk.

An unavailable button can reflect missing permission, unknown state, or an operation already in progress. A graceful action depends on guest support; lack of immediate response is not permission to force stop blindly.

The Node action restrictions diagnostic banner is for root administrators only. Its details are collapsed until opened. Hiding this banner from customers does not change capability checks or action restrictions. Each action’s permission, state, and availability checks still determine whether you can use it.

Request and verify an action

  1. Open Servers and select the server.
  2. Confirm its identity and current state on Overview.
  3. Save guest work and notify affected users.
  4. Select the intended power action.
  5. Read the confirmation, if shown.
  6. Enter the requested server identifier for a force action.
  7. Wait for the action result and refreshed state.
  8. Verify the expected guest or application state separately.

An accepted request is not a completed transition. Avoid repeated clicks while the operation is busy. A machine can be running while its application is unavailable.

When no power operation is recorded, the Panel shows no empty-history message. After a successful operation, a compact summary such as Last Start succeeded appears below the power buttons. The action label matches the recorded operation. It does not always say Start. Select the summary to expand its details. With a keyboard, press Tab to focus it, then Enter or Space. The details retain the matching task result, server state readback, and guest-boot limitation.

This result describes a past operation. It does not replace the current server-state badge. Succeeded confirms the matching task and its state readback, not the current server state, guest OS boot, or application health. Pending, failed, unknown, and status-read or recovery errors remain visible without expanding a successful result. No message does not prove that an action succeeded.

If the state does not change

Check Activity for a failure and Telemetry for freshness. If status is unknown or the node is unavailable, contact support with the action and time. Do not repeat force controls to resolve an offline host. Stopping runtime does not necessarily stop provider billing.

Console help · Recovery points

Type to search…

↑↓ navigate↵ selectEsc close