Description

Ensure that Compute Engine VM instances have the On host maintenance setting configured to Migrate VM instance (live migration), so that the instance is transparently moved to another host during Google-initiated infrastructure maintenance instead of being terminated and restarted.

Rationale

When On host maintenance is set to Terminate, the VM is stopped during a host maintenance event and must be manually or automatically restarted, resulting in downtime for the workload. Setting this option to Migrate allows Compute Engine to live-migrate the running instance to another host transparently, with no loss of in-memory state and minimal to no perceived downtime, which is generally preferable for most standard (non-preemptible, non-GPU/TPU-bound) production workloads.

Impact

  • Reduces or eliminates downtime during Google-initiated host maintenance events

  • Preserves in-memory instance state across the migration

  • Not available for certain machine types (such as those with GPUs or certain sole-tenant configurations), which may require Terminate instead

Default Value

On host maintenance defaults to Migrate VM instance for most standard machine types, but should be verified, particularly on custom or specialized machine configurations.

Pre-Requisites

  • IAM permissions to view and edit Compute Engine instance settings (e.g., Compute Instance Admin)

  • Confirmation that the instance's machine type supports live migration

Test Plan

  1. Sign in to the Google Cloud Console at https://console.cloud.google.com.

  2. Select the relevant project.

  3. Navigate to Compute Engine > VM instances and select the instance to review.

  4. Open the instance details and locate the Availability policy section.

  5. Verify that On host maintenance is set to Migrate VM instance.

  6. If it is set to Terminate VM instance and the machine type supports migration, follow the implementation steps.

Implementation Steps

  1. Sign in to the Google Cloud Console at https://console.cloud.google.com.

  2. Select the relevant project.

  3. Navigate to Compute Engine > VM instances and select the instance to update.

  4. Select Edit.

  5. Under Availability policy, set On host maintenance to Migrate VM instance.

  6. Click Save to apply the change.

  7. Confirm the instance details now reflect On host maintenance as Migrate VM instance.

Backout Plan

  1. Sign in to the Google Cloud Console at https://console.cloud.google.com.

  2. Select the relevant project.

  3. Navigate to Compute Engine > VM instances and select the affected instance.

  4. Select Edit.

  5. Under Availability policy, set On host maintenance back to Terminate VM instance.

  6. Click Save to apply the change.

Reference