Step 8: Failure Modes and Operations

A news service is most likely to fail at the moment it matters most. When a major story breaks, dozens of outlets publish within minutes and millions of users open the app at once. The notification service has its largest audience of the year at the same moment. Every load in the system peaks together, so the failures in this lesson tend to arrive together too.

This lesson covers six failures, how each one is detected, and how the design recovers. Two properties from earlier steps make most of the recoveries simple