What it does
Flagger is a Kubernetes progressive-delivery operator. It watches application revisions, gradually shifts traffic through a supported ingress controller or service mesh, and evaluates metrics or custom checks before promotion. The project documents canary, A/B and blue-green strategies, integration with monitoring systems, and webhooks for additional acceptance or load tests. It handles the rollout and validation stage after an application has been built; it is not a general-purpose build pipeline.
Flagger begins after a build exists: it stages a Kubernetes rollout instead of sending all traffic to a new revision at once. A Canary resource describes the analysis, and traffic routing plus metrics or custom checks inform the promotion decision.
A documented blue-green rehearsal
The official tutorial's example proceeds in three parts:
- Prepare a Kubernetes cluster at version 1.16 or newer and install Flagger with its Prometheus add-on.
- Deploy the sample application and create a Canary resource, with test traffic for the analysis.
- Update the sample image. Flagger observes the revision, evaluates the rollout and can promote or roll it back according to the configured checks.
This is a guide to the documented example; no cluster or workload was deployed.
The tool does not build source or replace CI tests. The team still needs compatible ingress or service-mesh routing and a metrics source; custom webhooks can add acceptance or load checks. Flagger belongs to the Flux project family but handles progressive delivery rather than Flux CD reconciliation. Its release page showed v1.45.0 as latest when checked.
Strengths
- Can gradually shift traffic and use metrics or conformance checks to guide promotion or rollback.
- Documents canary, A/B and blue-green deployment strategies with several Kubernetes routing integrations.
- Custom webhooks can run acceptance or load tests during an analysis.
Watch out for
- Requires a Kubernetes environment plus compatible traffic routing and metrics integrations.
- It covers progressive deployment, not source builds or the full CI pipeline.
Best for
- Kubernetes teams that want staged canary or blue-green releases
- Deployments that need metric-based promotion or rollback
- Teams adding acceptance or load-test checks to rollout analysis
Sources
- Flagger (Oct 25, 2026)
- Introduction | latest | Flagger (Oct 25, 2026)
- Blue/Green Deployments | Flagger (Oct 25, 2026)
- Metrics Analysis | latest | Flagger (Oct 25, 2026)
- Flagger (Oct 25, 2026)
- LICENSE | fluxcd/flagger (Oct 25, 2026)
- Releases | fluxcd/flagger (Oct 25, 2026)