Split-brain occurs when a network partition divides a cluster into disconnected groups that continue serving clients independently. Each group accepts input without seeing the other, creating conflicting state that may be impossible to reconcile after reconnection. section 12.2.1 section 12.2.5
Typical consequences and signals include:
- Requests accepted by one cluster remain invisible to the other, producing lost updates or phantom entries. section 12.2.5
- A peer that stops responding to periodic tick messages is treated as disconnected, although overload or bugs can cause the same symptom. section 12.2.5
- Processes can receive node-connectivity changes through :net_kernel.monitor_nodes/1. section 12.2.5
$ iex --sname node1@localhost iex(node1@localhost)1> :net_kernel.monitor_nodes(true)
Start a named Elixir node and enable node-monitor notifications before testing connections and disconnections. section 12.2.5
