Anti-Entropy
Summary: A background process that constantly compares data between replicas and copies any missing or outdated data to ensure that all replicas are in sync.
Sources: chapter5
Last updated: 2026-04-15
In leaderless replication, anti-entropy is a background process that constantly looks for differences in data between replicas (source: chapter5, p. 179).
Unlike read-repair, which only updates data that is being actively read, anti-entropy ensures that all data is eventually replicated, even if it is rarely accessed. It can be implemented using Merkle trees to efficiently compare large amounts of data between nodes.