A leader-based system depends on one simple rule: exactly one node is in charge at a time. Split-brain is what happens when that rule breaks and two or more nodes believe they are the leader.

In a healthy system, one node accepts writes and the others follow. That gives the system one clear history of changes. During a network partition, a slow failure check, or a bad failover, the old leader may keep accepting writes while another node is promoted to replace it. Now the system has two sources of truth.

The damage is rarely obvious right away. It usually shows up later, when the system tries to merge two histories that should never have existed at the same time. This chapter explains how split-brain happens and how real systems prevent it.