Split-brain
Split-brain is a distributed-system failure state in which disconnected nodes or groups independently act as authoritative and may make conflicting decisions or writes.
Elasticsearch in Action, Second Edition
A split-brain scenario is a distributed-system failure in which isolated node groups independently believe they are the authoritative cluster.
- How it is used
- Diagnoses cluster availability and data-consistency risks caused by node failure and lost network connectivity.
- How deep it goes
- Explains the failure sequence, independent request handling, quorum role, and three-node mitigation without covering implementation details.
- Examples
- Uses a two-node cluster where Node B loses contact with master Node A and independently becomes master.
- What you can do after
- Evaluate cluster topologies and identify why at least three master-eligible nodes reduce split-brain risk.
Read how Elasticsearch in Action, Second Edition explains split-brain →
Elixir in Action, Third Edition
Split-brain is a distributed-system condition in which mutually disconnected clusters independently continue serving clients as though they were authoritative.
- How it is used
- Frames network partitions as an availability and consistency hazard in distributed cluster design.
- How deep it goes
- Defines the failure mode, consequences, and connection-loss detection, while deferring partition-resolution strategies.
- Examples
- Demonstrates lost updates, phantom entries, conflicting data, and Elixir nodes reporting nodeup or nodedown events.
- What you can do after
- Diagnose partition symptoms with :net_kernel.monitor_nodes/1 and interpret nodeup or nodedown notifications.
Read how Elixir in Action, Third Edition explains split-brain →
Architecting an Apache Iceberg Lakehouse
A split-brain scenario is a distributed-system condition in which multiple systems independently believe they own the authoritative state of the same resource.
- How it is used
- Protects cross-environment Iceberg recovery by distinguishing inspection replicas from writable authoritative tables.
- How deep it goes
- Defines the failure mode, traces divergence and metadata corruption, and prescribes read-only registration or independent cloning.
- Examples
- Uses exported metadata.json, register_table, isolated test catalogs, and delayed cleanup during replication lag.
- What you can do after
- Operate cross-environment recovery without granting concurrent write ownership to shared table metadata.
Read how Architecting an Apache Iceberg Lakehouse explains split-brain →
Related concepts
- network partitionproduced by · A network partition is a failure condition in which disrupted connectivity prevents groups of otherwise operating components in a distributed system from communicating reliably.
- master-eligible nodeaddressed by · A master-eligible node is an Elasticsearch node assigned the master role and therefore able to vote in master elections and become the elected master.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.