What Is Raft and How Does It Work?

Distributed Services with Go — by Travis Jeffery (57 / 84)

👈 Chapter 8 Coordinate Your Services with Consensus | TOC | Implement Raft in Our Service 👉

Raft is a distributed consensus algorithm designed to be easily understood and implemented. It’s the consensus algorithm behind services like Etcd — the distributed key-value store that backs Kubernetes, Consul, and soon Kafka, whose team is migrating from ZooKeeper to Raft.[48] Because Raft is easy to understand and implement, developers have written many quality Raft libraries used in many projects and it’s become the most widely deployed consensus algorithm today.

Let’s talk about Raft’s leader election first and then talk about its replication, and that’ll transition into coding replication in our service.

Leader Election

A Raft cluster has one leader and the rest of the servers are followers. The leader maintains power by sending heartbeat requests to its followers, effectively saying: “I’m still here and I’m still the boss.” If the follower times out waiting for a heartbeat request from the leader, then the follower becomes a candidate and begins an election to decide the next leader. The candidate votes for itself and then requests votes from the followers. “The boss is gone! I’m the new boss, right?” If the candidate receives a majority of the votes, it becomes the leader, and it sends heartbeat requests to the followers to establish authority…