A subset of instances will be shut down and restarted as their host is updated. Affected instances will be unavailable for 15 to 30 minutes each, not for the full window. Owners of affected instances have been notified directly by email.

Compute operations via the Public API and console (create, start, stop, resize) may be delayed or fail during the window and should be retried afterwards. Storage volumes, IP assignments, DNS and billing are unaffected.