Resilience & Observability Testing — quiz

A team wants to test how their Order Service behaves when the Inventory Service goes down. They have never intentionally failed a dependency in a test environment. What is the safest starting approach?

• Run chaos experiments directly in production to observe real user-facing failures
• Start in a controlled test environment: define the steady-state behaviour first, then inject the failure (e.g. stopping the Inventory Service container), measure the impact, and verify the Order Service degrades gracefully
• Simulate the failure with a code-level mock that returns errors rather than actually stopping the service
• Add a feature flag to the Inventory Service that forces it to return errors on demand