In the previous chapter, we explored the fundamental challenges of distributed systems: partial failures, unreliable networks, and unsynchronized clocks.

Among all these challenges, network partitions are perhaps the most disruptive. A partition does not just slow things down or cause occasional failures. It splits your system in two, creating isolated islands of nodes that cannot communicate with each other.

A network partition is not a theoretical concern. It happens in production, often at the worst possible time. Cloud providers experience them. Data centers experience them. Even a misconfigured firewall rule or a failed switch can create a partition that lasts seconds, minutes, or hours. When it happens, your system must make hard choices: stop serving requests to maintain consistency, or continue operating and risk data divergence.

Understanding network partitions is essential because they force your hand. They reveal the true design of your system, exposing whether you prioritized consistency or availability, and whether your failure handling actually works.

What Is a Network Partition?