Canary with Argo Rollouts
Progressive delivery driven by real metrics.
0 of 6 steps
What you will end up with
A metric-driven canary
Automated analysis
Automatic rollback
Build steps
0/6Apply the Argo Rollouts install manifest along with its kubectl plugin. This swaps in a smarter traffic conductor to replace the basic Deployment controller.
Change kind: Deployment to kind: Rollout in your YAML, keeping the same pod template underneath. Same car, just handing the keys to a driver who knows how to merge traffic gradually.
Add a list of steps like setWeight: 20, then a pause, then setWeight: 50. Each step is a rung on a ladder you climb one at a time, never a single leap to the top.
Add an AnalysisTemplate that queries Prometheus for error rate or latency, then reference it from the canary steps. This hands the ladder-climbing decision to a sensor instead of a person’s gut feeling.
Update the image tag in the Rollout spec and apply it; Argo Rollouts automatically starts stepping through the canary plan on its own. That single image bump kicks off the whole climb.
Watch the rollout’s status and confirm it aborts and reverts on its own if the analysis template flags a bad metric. The ladder should retract by itself the moment a rung feels shaky.
Before you start