A newer version of Hazelcast Platform is available.

View latest

Network Partitioning

Split-Brain Syndrome

In general, network partitioning is a network failure that causes the members to split into multiple groups such that a member in a group cannot communicate with members in other groups. In a partition scenario, all sides of the original cluster operate independently assuming members in other sides are failed. Network partitioning is also called as Split-Brain Syndrome.

Even though this communication failure is called as network partitioning, in practice a process or an entire OS that’s suspending/pausing very long can cause communication interruptions. If these interruptions take long enough time to assume that the other side is crashed, the cluster splits into multiple partitions and they start operating independently. That’s why any communication failure/interruption long enough can be classified as network partitioning.

Moreover, communication failures don’t have to be symmetrical. A network failure can interrupt only one side of the channel or a suspended process/member may not even observe the rest as crashed. That kind of network partitioning can be called as partial network partitioning. See Partial Network Partitions section for more info.