Automated verification and rollback
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
let metrics, not people at 2 a.m., decide whether to continue or revert.
Planned topics
- Pick a few guard metrics: error rate, latency percentile, saturation, and one business metric such as orders completed.
- Compare canary against baseline over the same window, with a tolerance that accounts for noise.
- Automate the abort: a failed check reverts traffic without waiting for a human.
- Rehearse rollback; an untested rollback is a hope, not a plan.
Draft outline details
Demonstration
Analysis rule in words: abort if canary 5xx rate exceeds baseline by more than 0.5 points for 5 minutes, or p99 is more than 20% worse, or order completion drops below baseline minus 2%. Thresholds are examples to tune.
Check your understanding
Why include a business metric next to technical ones?
Show answer
A change can return 200s yet break checkout outcomes.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.