network-partition · shipping · also Network partitioning

Network partition

A network partition is a communication failure that prevents message exchange between groups of nodes that may still be running.

Related Observatory article: Network partition on hesela.com

Scope: A lost heartbeat does not prove that a peer has stopped or lost storage access. The communication failure and the enforcement of resource exclusion are separate observations. [1]

Formal boundary: The CAP model permits arbitrary message loss between components. It cannot guarantee both linearizable reads and writes and eventual completion of every read/write request received by a non-failing node across all such executions. This availability definition is not an uptime percentage or a latency deadline. [2]

Relations

related
Split brain, I/O fencing, Failure detector, Quorum intersection, Linearizability

Citations

  1. ClusterLabs: Pacemaker Explained 3.0, Fencing, sections 8.1-8.3 and 8.11 (accessed 2026-11-09)
  2. Gilbert and Lynch (2002): Brewer's Conjecture and the Feasibility of Consistent, Available, Partition-Tolerant Web Services (accessed 2026-11-09)
Sources & caveats
  • Source review of Pacemaker 3.0 and Gilbert and Lynch sections 2.1-2.3 and 3.1, accessed 2026-11-09. No cluster fault-injection experiment.

JSON: /data/terms/network-partition.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Network partition. Hesela. 2026-11-09. https://hesela.dev/terms/network-partition/