Raft Consensus Algorithm Visualiser

The Raft consensus algorithm, as specified in Ongaro & Ousterhout (2014), is designed for understandability — yet most available implementations simplify or omit portions of the specification in favour of demonstration. This project takes the opposite approach: a precise implementation of the extended paper, conducted alongside the MIT 6.5840 Distributed Systems course (formerly 6.824), with a visualisation layer for real-time inspection of cluster behaviour.

The system comprises two components:

• Raft cluster — Go nodes communicating via gRPC, following the paper's RPC definitions (RequestVote, AppendEntries, InstallSnapshot). Instrumented with OpenTelemetry Go SDK for distributed tracing across all inter-node RPCs.
• Visualiser — React frontend connected to the cluster over gRPC, rendering cluster state transitions as they occur. Instrumented with OpenTelemetry JS SDK. Telemetry from both layers is aggregated through an OTEL Collector.

Roadmap

Implementation follows the paper's structure. Each phase corresponds to a section of the specification.

Raft Cluster (Go + gRPC + OTEL)

• Leader election — term management, RequestVote RPC, randomised election timeouts, split vote resolution
• Log replication — AppendEntries RPC, log consistency check, commitment by majority
• Safety — election restriction (§5.4.1), commitment rules for entries from prior terms (§5.4.2)
• Persistence — stable storage of currentTerm, votedFor, and log entries across restarts [As a cluster seldomly been restart in this practice project then retain to be a could have]
• Log compaction — InstallSnapshot RPC, state machine snapshotting (§7) [retain to be a could have requirement]
• Cluster membership changes — joint consensus for configuration transitions (§6) [retain to be a could have requirement]
• OTEL instrumentation — tracing across all RPCs and state transitions

Visualisation (React + OTEL)

• Cluster topology — node states (follower, candidate, leader), current term, leader identity
• Log replication — per-node log state, commit index progression
• Election sequence — vote requests, grants, term transitions, timeout events
• OTEL JS SDK integration — end-to-end trace correlation with the cluster layer

References

• In Search of an Understandable Consensus Algorithm (Extended Version) — Ongaro & Ousterhout, 2014
• MIT 6.5840 Distributed Systems