Azure Front Door: incompatible metadata crashes the edge network for over eight hours

2025-10-29 · Publisher : Microsoft · Domain : Cloud & DevOps

On 29 October 2025, customer configuration changes generate incompatible metadata that exposes a bug in Azure Front Door. The Azure portal, Microsoft 365 and many services are affected for about 8.5 hours.

What changes

  • From 29 October 15:41 UTC to 30 October 00:05 UTC: about 8.5 hours of incident.
  • Cause: incompatible metadata between two control plane versions, which triggered a latent data plane bug.
  • Azure portal, SQL Database, App Service, Microsoft 365, Dynamics 365 and Copilot affected.
  • Recovery by manual removal of corrupted configurations, then phased traffic rebalancing.

### Background On 29 October 2025, from 15:41 UTC to 00:05 UTC on 30 October (about **8.5 hours**), **Azure Front Door** (AFD), Microsoft's global edge delivery and load-balancing service, suffered a major failure. Microsoft's post-incident review (PIR) explains it as follows: a sequence of **customer configuration changes**, processed by two different versions of the control plane, generated **incompatible metadata**. It exposed a **latent bug in the data plane**, which crashed during asynchronous processing. The defect had gone undetected because not all features are validated across control plane versions. ### Key facts - Services affected: the Azure portal, SQL Database, App Service, Active Directory B2C, but also Microsoft 365, Dynamics 365, Copilot and customer support; customers saw timeout and DNS resolution errors. - The configuration protection system blocked new deployments at 15:43 UTC; the faulty changes were identified at 16:15. - Manual removal of corrupted configurations from the "last known good" snapshot from 17:10, deployment of the corrected configuration from 17:40. - Phased recovery with traffic rebalancing, completed at 00:05 UTC on 30 October. ### Why it matters Another major failure of a global service, less than ten days after AWS's, is a reminder that shared entry points are critical dependencies, including for Microsoft's own services. For IT leaders: plan a bypass of the edge service (direct access to the origin, fallback DNS), and include identity and portal services in crisis scenarios. ### Takeaway Azure's global entry point fell to a compatibility defect between versions of its own control plane.

Links

Tags : Incident, Azure, Edge, Hyperscaler