Question

Your team runs a customer-facing web API on Azure Kubernetes Service, deployed through a multi-stage Azure Pipeline. Leadership wants new features exposed to only 10% of live traffic initially, then progressively increased, and wants the ability to instantly disable a feature at runtime WITHOUT triggering a new deployment or pod restart. Which approach best meets all these requirements?

Accepted Answer

Use Azure App Configuration feature flags with a percentage filter, evaluated by the application at request time

Answer

Configure a blue-green deployment strategy that swaps the entire environment once smoke tests pass

Answer

Add a deployment gate in the release pipeline that queries Application Insights before promoting to production

Answer

Set an environment variable in the AKS deployment manifest and redeploy whenever a feature must change