Rolling updates are fine until a bad release reaches every user before you notice. Blue-green and canary deployments give you a safety valve: expose the new version to a slice of traffic (or a parallel environment) and promote only when it looks healthy.
Blue-green
Run two identical environments, 'blue' (current) and 'green' (new). Cut traffic over by flipping a Service selector; roll back instantly by flipping it back.
apiVersion: v1
kind: Service
metadata: { name: web }
spec:
selector: { app: web, slot: green } # flip to blue to roll back
ports: [{ port: 80, targetPort: 3000 }]Canary with weighted traffic
A canary shifts a small percentage of traffic to the new version and increases it as metrics stay green. With a mesh like Istio this is a weight change.
http:
- route:
- destination: { host: web, subset: v1 }
weight: 90
- destination: { host: web, subset: v2 }
weight: 10- Blue-green: instant switch/rollback, but double the resources briefly
- Canary: gradual, metric-driven, cheaper, but slower to fully roll out
- Automate promotion on SLOs (error rate, latency) with Argo Rollouts or Flagger
The goal is not zero incidents - it is small blast radius and instant rollback.
Pick blue-green for simple, fast cutovers and canary when you want evidence from real traffic before committing.