Glossary · System coordination, integration and orchestration
Split-brain condition
Also known as: Split brain
German: Split-Brain-Zustand
In redundant and distributed systems, a split-brain condition is a fault state in which the nodes of a cluster or redundant pair lose communication with each other and more than one node acts as the active primary, leading to conflicting decisions, outputs or data.
- System integration
In one sentence
A split-brain condition occurs when redundant nodes lose contact and more than one acts as primary, producing conflicting outputs or data.
Example
After the synchronization link between two redundant SCADA servers fails, both become active and send conflicting setpoints to the same controller.
How it applies
- Engineering: Prevent split brain with redundant synchronization links, quorum or witness mechanisms and fencing (forcibly isolating the node that must not be active). A redundant pair without a tie-breaker cannot distinguish "partner failed" from "link failed".
- Commissioning: Test link failures explicitly, not only node failures; split brain typically appears only in this case.
- Operation: Alarm on loss of synchronization, and define the manual procedure to resolve split brain and reconcile data.
- Documentation: The system administration documentation should describe the redundancy architecture, how split brain is prevented and the recovery procedure. Operators need to know which alarms indicate the condition.
Split brain vs. failover
A failover or Switchover is the intended transfer of the active role from one node to another. Split brain is the unintended result when a failover happens even though the original primary is still active.