Exercises › Progressive Delivery Language › Rollback Strategy Vocabulary

Rollback Strategy Vocabulary

5 questions · Progressive Delivery Language

1. An incident commander states: "The error rate exceeded 1% — rollback was triggered automatically." What type of system produces this behaviour?

A manual runbook that the on-call engineer follows when error rates exceed defined thresholds An automated progressive delivery controller (such as Flagger or Argo Rollouts) that continuously monitors metrics and initiates rollback when a configured threshold is breached A circuit breaker on the client side that stops sending requests to the failing service

2. A team debates whether to roll back or roll forward after a bad deployment. What does "roll forward" mean in this context?

Running the deployment pipeline again to deploy the same version that failed, in case the issue was a transient infrastructure problem Quickly deploying a new fixed version (v3) to replace the broken version (v2) rather than reverting to the previous stable version (v1) Forwarding the incident to the next on-call engineer if the current one cannot resolve the issue within 30 minutes

3. An engineer says: "We can roll back this feature in seconds using the feature flag — no deployment needed." How does feature flag rollback differ from a deployment rollback?

Feature flag rollback is less reliable than deployment rollback because flags can have propagation delays Feature flag rollback disables the feature by flipping a flag (taking effect in milliseconds to seconds, no deployment pipeline required); deployment rollback reverts to a previous container image or artifact (taking minutes, requires CI/CD pipeline execution) Feature flag rollback only works for frontend features; deployment rollback is required for backend changes

4. A team documents their rollback triggers in the deployment runbook. Which set of conditions represents appropriate automated rollback triggers?

Customer complaints received on Twitter, team lead approval withheld, or any deployment taking longer than expected Error rate exceeds baseline by more than 0.5%, p99 latency increases by more than 20% vs. stable version, or success rate drops below 99% Any alert firing in the monitoring system during the deployment window, regardless of severity

5. After an automatic rollback, the incident commander communicates to stakeholders: "We detected elevated error rates on the new deployment and rolled back to the stable version. Users experienced degraded service for approximately 3 minutes." What term describes this 3-minute window?

The deployment window — the scheduled time allocated for the deployment to complete The blast radius duration — how long the impact of the bad deployment lasted before rollback restored normal service The detection latency — the time it takes the monitoring system to first detect an anomaly

Exercise complete!

out of 5 questions

← Back to Progressive Delivery Language All exercises