Azure CDN outage raises reliability flags

Microsoft’s Azure content delivery network failed in two separate October 2025 incidents, and Microsoft’s own writeups trace both outages to control-plane mistakes that spread bad changes across its edge network. (azure.status.microsoft) A content delivery network is the layer that serves websites and apps from edge locations close to users. Microsoft says Azure Front Door and Azure Content Delivery Network use globally distributed edge sites and also sit in front of Microsoft services such as the Azure portal. (techcommunity.microsoft.com) The weak point in both incidents was the control plane, the management system that creates and pushes configuration, not the traffic-serving data plane alone. Microsoft said the control plane generates metadata for create, update, delete, purge, and web application firewall operations on Azure Front Door and Azure Content Delivery Network profiles. (azure.status.microsoft) In the larger outage, Microsoft said customers and Microsoft services saw connection timeouts and Domain Name System resolution issues from 15:41 Coordinated Universal Time on October 29 until 00:05 Coordinated Universal Time on October 30, 2025. Recovery began around 18:30 Coordinated Universal Time but latency remained elevated until the system stabilized. (azure.status.microsoft) Microsoft listed Azure Active Directory B2C, Azure App Service, Azure Portal, Azure SQL Database, Azure Static Web Apps, Microsoft 365, Microsoft Entra ID, Microsoft Purview, and customer support case systems among the affected services. The company said a sequence of valid customer configuration changes across two control-plane build versions generated incompatible metadata that exposed a latent bug in the data plane. (azure.status.microsoft) The earlier incident hit on October 9, 2025. Microsoft said a manual cleanup of stuck tenant metadata bypassed its configuration protection layer, letting incompatible metadata move beyond canary edge sites after a control-plane defect had created that metadata on October 7. (techcommunity.microsoft.com) Microsoft’s December 18, 2025 engineering post said Azure Front Door runs through more than 210 global and local points of presence, and that dense sharing across many tenants can turn a bad configuration into a broad outage if it is not contained quickly. The company said the October incidents pushed it to accelerate work on tenant isolation and resilience hardening. (techcommunity.microsoft.com) The reliability concern is not limited to one product line. On February 2, 2026, Microsoft published another post-incident review for a separate multi-region control-plane failure that disrupted virtual machines, Azure Kubernetes Service, Azure DevOps, and GitHub Actions after a remediation workflow changed access settings on Microsoft-managed storage accounts. (azure.status.microsoft) Dawn Liphardt’s April 12, 2026 post ties the October 9 outage to a software defect in the Azure Front Door control plane and says a later recovery step removed a load-balancing configuration value, producing a second problem during restoration. That account lines up in broad outline with Microsoft’s own description of one October incident driven by a control-plane defect and another driven by configuration propagation. (dawnliphardt.com) Microsoft’s public record now shows a pattern: when the management layer for Azure Front Door fails, the fallout can reach Microsoft’s own portals and customer applications at the same time. The company says it is rebuilding protections around configuration rollout, validation, and tenant isolation to keep the next bad change from traveling so far. (techcommunity.microsoft.com)