Two DCs, 3 nodes each. Click 'Toggle partition' to break the link between them.

What you're seeing

CAP: under a network partition, you must choose between consistency and availability.

AP (Cassandra): both sides keep accepting writes. They diverge during the partition; reconcile (last-write-wins, vector clocks) when partition heals. Available; not consistent.

CP (etcd, ZooKeeper): only the side with majority accepts writes. The minority side returns errors. Consistent; not fully available.

★ KEY TAKEAWAY
Under network partition you must choose: AP (keep accepting writes, diverge) or CP (block writes on the minority side, stay consistent).
▶ WHAT TO TRY
  • Click Toggle partition to split the cluster.
  • Switch between AP and CP modes — see how writes behave differently.
  • Cassandra is AP; etcd/ZooKeeper are CP.
← Back to Distributed Systems