What Is Canary Deployment?

The name comes from the "canary in the coal mine" practice. In a canary deployment, the new version is deployed alongside the stable version. A load balancer or traffic router directs a small percentage (typically 1-5%) of traffic to the canary. If error rates, latency, or business metrics remain healthy, traffic is gradually shifted (10%, 25%, 50%, 100%). If anomalies are detected, traffic is immediately routed back to the stable version.

Canary deployments require robust observability. You need real-time dashboards comparing canary metrics against the baseline: HTTP error rates, response latency percentiles (p50, p95, p99), CPU/memory usage, and business KPIs (conversion rate, cart abandonment). Automated analysis can even trigger rollback without human intervention when metrics exceed thresholds.

From a QA perspective, canary deployments are a shift-right testing technique. They expose code to real traffic patterns that synthetic tests cannot replicate. The QA role shifts from "certify before release" to "monitor and validate during release," requiring skills in observability tooling, metric interpretation, and incident response.