Failure Modes in Distributed Systems

Microservices rarely fail as one piece. One service may be down, slow, unreachable, or returning bad responses while the rest of the system keeps running.

That makes failure harder to reason about. A service-to-service call must be designed for the moment a dependency stops cooperating, not just the happy path.

This chapter covers common distributed system failures: partial failure, gray failure, network partitions, blast radius, and the key question every design should ask: what happens when this call fails?