A gossip protocol spreads information through a cluster by having nodes periodically exchange state with a small number of peers.

Instead of one node broadcasting an update to every other node, each node shares what it knows with a few peers. Those peers do the same in later rounds. Information spreads through many paths, so the system can tolerate dropped messages, failed nodes, and partial network problems.

Gossip is useful for data that can be eventually consistent:

Gossip complements consensus. Gossip tells nodes about state over time. Consensus makes a group agree on one decision before moving forward.