DevOps Practitioner · Module 6: Safe delivery and release engineering · Lab
Show Canary Delivery and Roll Back on Failure
55 min hands-on · Advanced
A local KinD or k3d cluster; a demo service releasable in two versions with a traffic split (two Deployments with weighted routing or replica-share canary).
Local guide: run the steps below on your own machine in order, then check the validation list.
Objectives
- Route a small share of live traffic to the new version
- Compare canary against baseline with traffic evidence
- Roll back on failure and drain the canary
Step 1
Split the traffic
Deploy the new (faulty) version alongside the current one with a small traffic share. Generate load and record both versions' error rates separately.
Step 2
Read the comparison
Quote the canary share failing while the majority serves. Name the rollback criterion the numbers trigger (error budget, threshold) before acting.
Step 3
Roll back and drain
Roll back through the recorded path, drain the canary share to zero, and verify all traffic clean on the previous version. Record times for detection, decision and restore.
How to confirm it worked
- Canary errors quoted against a clean baseline
- Rollback criterion named before the action
- Canary drained with all traffic verified clean
- Detection, decision and restore times recorded