Q: Your engineering squads perform risky 'all-at-once' rolling deployments, causing production outages during unexpected regressions. How do you design an enterprise-wide Progressive Delivery standard using Argo Rollouts, automated Prometheus analysis templates, and automated rollbacks?

Create a reusable `AnalysisTemplate` measuring HTTP 5xx error rate and p99 latency over 5-minute evaluation windows.

apiVersion: argoproj.io/v1alpha1
kind: AnalysisTemplate
metadata:
  name: service-success-rate
  namespace: platform-governance
spec:
  metrics:
    - name: success-rate
      interval: 1m
      successCondition: result[0] >= 0.99
      failureLimit: 2
      provider:
        prometheus:
          address: http://prometheus.monitoring:9090
          query: |
            sum(rate(http_requests_total{status!~"5.*", app="{{args.service-name}}"}[2m]))
            /
            sum(rate(http_requests_total{app="{{args.service-name}}"}[2m]))

Replace `kind: Deployment` with `kind: Rollout` in the platform's Golden Path Helm chart. Configure progressive canary steps (10% -> 25% -> 50% -> 100%) with automated pause intervals.

apiVersion: argoproj.io/v1alpha1
kind: Rollout
spec:
  strategy:
    canary:
      analysis:
        templates:
          - templateName: service-success-rate
        args:
          - name: service-name
            value: checkout-api
      steps:
        - setWeight: 10
        - pause: { duration: 5m }
        - setWeight: 50
        - pause: { duration: 10m }

If Prometheus reports `success-rate < 0.99`, Argo Rollouts immediately shifts 100% of ingress traffic back to the stable ReplicaSet and alerts the squad via Slack.