Spinnaker — Advanced Deployment Strategies

Rolling update in Kubernetes is the foundation. But for critical services, we need more: canary deployment with automated analysis, blue-green for zero-downtime cutover, automated rollback based on metrics.

Why Spinnaker¶

Netflix created Spinnaker to manage thousands of deployments per day. Multi-cloud support (AWS, GCP, Azure, Kubernetes), pipelines with manual approval gates, and most importantly: Automated Canary Analysis (ACA).

Canary Deployment with Kayenta¶

Kayenta is a Spinnaker module for automated canary analysis. It compares metrics of the canary version with the baseline (production). Error rate higher? Latency worse? Automatic rollback. Metrics OK? Gradually increase traffic.

Canary Analysis Config:
  metrics:
    - name: error_rate
      query: rate(http_errors_total[5m])
      threshold: marginal=5%, pass=1%
    - name: latency_p99
      query: histogram_quantile(0.99, ...)
      threshold: marginal=10%, pass=5%
  lifetime: 30m
  score_threshold: 70

Blue-Green Deployment¶

Two identical production environments. Green (new version) is deployed and tested, then the load balancer switches. Problem? Instant rollback by switching back to blue.

Spinnaker vs. ArgoCD¶

ArgoCD is simpler and Kubernetes-native. Spinnaker is more complex but offers multi-cloud, canary analysis, and sophisticated pipelines. For pure Kubernetes projects, we choose ArgoCD. For multi-cloud and advanced deployment strategies, Spinnaker.

Advanced Deployment Strategies Reduce Risk¶

Canary deployment with automated analysis catches problems before they affect all users. For critical services, it’s an investment that pays off.