// the find
fluxcd/flagger
Progressive delivery Kubernetes operator (Canary, A/B Testing and Blue/Green deployments)
Flagger is a Kubernetes operator that automates progressive delivery — canary releases, A/B testing, and blue/green deployments — by watching your Deployments and gradually shifting traffic based on real metrics. It's for platform teams that want automated rollback without writing their own controllers. CNCF graduated, part of the Flux ecosystem.
Works with a wide range of ingress controllers and service meshes (Istio, Linkerd, NGINX, Contour, Gateway API) without requiring you to pick one ecosystem lock-in. The Canary CRD also tracks ConfigMap and Secret changes, so a bad config rotation triggers the same canary analysis as a bad image — most competitors miss this. Webhook system is genuinely useful: pre-rollout gates, in-flight load testing, and post-promotion hooks all compose cleanly. Metric analysis is pluggable via MetricTemplate CRDs backed by any Prometheus-compatible source, so you're not stuck with Flagger's built-in checks.
The feature matrix shows real gaps: A/B testing only works with Istio on the service mesh side, and traffic mirroring is basically Istio-only too — if you're on Linkerd or Gateway API, you get a subset of the selling points. The project is still on its own controller-runtime predating kubebuilder norms, and the roadmap item to migrate has been sitting there without a clear timeline. No native support for non-HTTP workloads — if you're running gRPC-heavy or raw TCP services, you'll hit walls fast. Debugging a stuck canary means parsing Flagger operator logs and cross-referencing service mesh telemetry simultaneously, which is painful without strong Grafana/Prometheus already in place.