Anatomy of Raft

A complete, granular dissection of Ongaro and Ousterhout (2014). Five servers lie on a 3D table as rows, the way the paper draws its logs: each has its state, its election timer, its log, its state machine and, on a leader, nextIndex and matchIndex, with messages as links that draw on. A seeded run elects a leader, replicates and commits commands, crashes the leader holding an unsent entry, elects another, repairs the old leader's log and heals a cut-off follower. The paper's Figure 7 (six follower logs) and Figure 8 (why old terms are not counted) are rebuilt from real blocks and run through the engine, the five properties of Figure 3 are checked on 300 random executions, and joint consensus is enumerated exhaustively. A 28-step article cites every number to its sentence, notes where the paper contradicts itself, and ends with a research desk of three pre-registered experiments: Figure 16 reproduced, the dissertation's closed-form split-vote model checked against simulation, and where the timing requirement breaks.

What is real. The run is a seeded discrete-event simulation of the rules of the paper's Figure 2, computed in the page by engine.js. test.mjs checks it against Figure 7 (rejections per follower counted by a separate walk down the logs), Figure 8 (panels d and e recomputed), the five properties of Figure 3 on 300 random executions with crashes, partitions, loss and duplication, two deliberately broken rules, the dissertation's closed form against the numbers its text quotes, an exhaustive enumeration of configuration changes against a pigeonhole condition, and the stored research results recomputed from their seeds. Every paper number is a fact with a verbatim quote from the paper, the dissertation or the source code.