Microsoft’s West US Azure region suffered an outage lasting almost five hours after routine network maintenance mistakenly removed IP routes between a datacentre and Microsoft’s wide-area network. The disruption affected 27 Azure services and prevented traffic entering or leaving the Californian region, underlining how maintenance errors can affect supposedly resilient cloud infrastructure.

The incident began at 14:44 UTC on 23 July, when a bug in Microsoft’s maintenance-request conversion system incorrectly included additional devices in the work. Microsoft began rolling back changes at 17:45 UTC; its wide-area network recovered by 18:26, and all affected services were restored by 19:41.