Q36 of 38 · CI/CD & DevOps
How do you build progressive delivery pipelines with automated quality gates?
Short answer
Short answer: Progressive delivery incrementally shifts traffic to a new release (1% → 10% → 100%) and uses automated observability checks at each increment as a quality gate. QA's role is defining the pass/fail criteria and ensuring the rollback path is tested before it is needed.
Detail
Tools like Argo Rollouts, Flagger, and AWS CodeDeploy support canary and blue-green strategies natively. The pipeline declares traffic weights and an analysis template that must pass before each increment proceeds.
The analysis template queries real metrics: error rate from Prometheus, p99 latency from Datadog, business KPIs from a custom metric. If any metric breaches its threshold during the canary window, the deployment rolls back automatically without human intervention.
QA's responsibilities in this model: 1. Define the thresholds in the analysis template — not just devops. QA understands acceptable error rates and latency budgets from the performance test history. 2. Write synthetic canary checks — short smoke tests that run specifically against the canary pods, not just through the load balancer. 3. Test the rollback path in staging — deliberately breach a threshold and verify the automated rollback fires and post-rollback smoke passes.