Group Replication
MySQL's built-in solution for synchronous, consensus-based replication — the foundation of InnoDB Cluster. A group of servers agrees via a Paxos-like protocol on which transactions to commit, giving automatic failover and strong consistency.
Mechanic: when a transaction is to be committed, it's sent to the group, which votes via consensus (the Group Communication System, based on a Paxos variant) on whether to accept it. A majority (quorum) is required → the system survives a minority of nodes failing. Modes: single-primary (one write node, the rest read replicas, with automatic election of a new primary on failure) or multi-primary (all can write, with conflict detection). Win over traditional asynchronous replication: no data loss on failover, automatic membership management. Competes with Galera Cluster (similar idea, different implementation). The core of MySQL's high-availability offering. Related to Galera Cluster and quorum.