Can Isabelle help to prove the correctness of the Raft or Paxos consensus algorithm? Is there a formal proof somewhere?
2 Answers
This entry in the Archive of Formal Proofs might be useful: https://www.isa-afp.org/entries/DiskPaxos.html

- 535
- 2
- 7
Isabelle/HOL (and similar theorem provers) are well-fit (and regularly used) for the verification of distributed protocols like Raft and Paxos.
Usually, the formalization will be about the systems model (messages or shared state, faults?), an inductive definition of the state space, a description of the temporal correctness properties and the relevant (inductive) invariants to prove the properties.
A tutorial video can be found here: https://www.youtube.com/watch?v=Uav5jWHNghY
As lsf37 has pointed out, there is a verification of Disk Paxos by Jaskelioff and Merz in the Archive of Formal Proofs from 2005.
An Isabelle/HOL formalization of classic Paxos has been done by Küfner et al. in “Formal Verification of Distributed Algorithms: From Pseudo Code to Checked Proofs“ (2012, more details in Küfner's dissertation.)
Moreover, there is an Isabelle proof of the theoretical impossibility of consensus in the asynchronous setting with one faulty process (the so-called “FLP result”) by Bisping et al. from 2016.
For Raft, there is a Coq/Verdi formalization

- 671
- 4
- 11