[][src]Crate paxakos


Paxakos is a pure Rust implementation of a distributed consensus algorithm based on Leslie Lamport's Paxos. It enables distributed systems to consistently modify shared state across their network, even in the presence of failures.


In order to use Paxakos, the traits LogEntry, State, NodeInfo and Communicator need to be implemented. The first two describe what state will be replicated across the network and which operations on it are available. The latter describe the nodes in your network and how to communicate between them.

Below are two partial implementations of LogEntry and State. The other two traits are more abstract and won't be illustrated here.

use std::collections::HashSet;

use paxakos::{LogEntry, State};
use uuid::Uuid;

#[derive(Clone, Copy, Debug)]
pub enum CalcOp {
    Add(f64, Uuid),
    Div(f64, Uuid),
    Mul(f64, Uuid),
    Sub(f64, Uuid),

impl LogEntry for CalcOp {
    type Id = Uuid;

    fn id(&self) -> Self::Id {
        match self {
            CalcOp::Add(_, id)
            | CalcOp::Div(_, id)
            | CalcOp::Mul(_, id)
            | CalcOp::Sub(_, id) => {

#[derive(Clone, Debug)]
pub struct CalcState {
    applied: HashSet<Uuid>,
    value: f64,

impl State for CalcState {
    type Context = ();

    type LogEntry = CalcOp;
    type Outcome = f64;
    type Event = f64;

    fn apply(
        &mut self,
        log_entry: &Self::LogEntry,
        _context: &mut (),
    ) -> (Self::Outcome, Self::Event) {
        if self.applied.insert(log_entry.id()) {
            match log_entry {
                CalcOp::Add(v, _) => {
                    self.value += v;
                CalcOp::Div(v, _) => {
                    self.value /= v;
                CalcOp::Mul(v, _) => {
                    self.value *= v;
                CalcOp::Sub(v, _) => {
                    self.value -= v;

        (self.value, self.value)

Working example

Here is an excerpt of the chat example, which gives a more complete picture of how the API is used once the necessary traits are implemented.

//! This example is for illustrative purposes only. Basing a chat protocol on
//! consensus is a bad idea, but makes a neat example.

use paxakos::append::AppendArgs;
use paxakos::prototyping::{DirectCommunicator, PrototypingNode, RetryIndefinitely};
use paxakos::{LogEntry, Node, NodeBuilder, NodeHandle, NodeInfo, RoundNum, State};

type ChatCommunicator = DirectCommunicator<ChatState, u64, u32>;

let node_a = PrototypingNode::new();
let node_b = PrototypingNode::new();
let node_c = PrototypingNode::new();

let nodes = vec![node_a, node_b, node_c];

let communicator = ChatCommunicator::new();

let node_a = spawn_node(node_a, nodes.clone(), communicator.clone());
let node_b = spawn_node(node_b, nodes.clone(), communicator.clone());
let node_c = spawn_node(node_c, nodes, communicator);

futures::executor::block_on(async move {
    let _ = node_a
        .append(msg("Alice", "Oh, hey guys"), always_retry())

// Because Bob and Charlie reply without synchronization, either may reply
// first. However, all participants will observe the same person replying
// first.
let b = std::thread::spawn(|| {
futures::executor::block_on(async move {
        let _ = node_b
            .append(msg("Bob", "Hi Alice, long time no see!"), always_retry())
let c = std::thread::spawn(|| {
    futures::executor::block_on(async move {
        let _ = node_c
            .append(msg("Charlie", "Hi Alice, how are you?"), always_retry())

// Let's wait for the appends to go through.

// It is guaranteed that all messages above have been appended to the shared log
// at this point. However, one node may not know about it yet and the others may
// not have gotten a chance to apply it to their state. Let's give them a chance
// to do that.

// Graceful shutdown is possible (see `Node::shut_down`) but is too involved for
// this example.

fn spawn_node(
    node_info: PrototypingNode,
    all_nodes: Vec<PrototypingNode>,
    communicator: ChatCommunicator,
) -> NodeHandle<ChatState, u64, u32> {
    let (send, recv) = futures::channel::oneshot::channel();

    std::thread::spawn(move || {
        let (handler, mut node) = futures::executor::block_on(
                .with_initial_state(ChatState::new(node_info.id(), all_nodes))


        communicator.register(node_info.id(), handler);

        futures::executor::block_on(futures::future::poll_fn(|cx| {
            let _ = node.poll_events(cx);



fn msg(sender: &str, message: &str) -> ChatMessage {
    ChatMessage {
        id: Uuid::new_v4(),
        sender: sender.to_string(),
        message: message.to_string(),

fn always_retry<R: RoundNum>() -> AppendArgs<R> {
    AppendArgs {
        retry_policy: Box::new(RetryIndefinitely::without_pausing()),

#[derive(Clone, Debug)]
pub struct ChatMessage {
    id: Uuid,
    sender: String,
    message: String,

impl LogEntry for ChatMessage {
    type Id = Uuid;

    fn id(&self) -> Self::Id {

#[derive(Clone, Debug)]
pub struct ChatState {
    node_id: usize,
    nodes: Vec<PrototypingNode>,

impl ChatState {
    pub fn new(node_id: usize, nodes: Vec<PrototypingNode>) -> Self {
        Self { node_id, nodes }

impl State for ChatState {
    type Context = ();

    type LogEntry = ChatMessage;
    type Outcome = ();
    type Event = ();

    type Node = PrototypingNode;

    fn apply(
        &mut self,
        log_entry: &Self::LogEntry,
        _context: &mut (),
    ) -> (Self::Outcome, Self::Event) {
        let own_node_id = format!("{:X}", self.node_id + 10);

            "[{}] -- {}: {}",
            own_node_id, log_entry.sender, log_entry.message

        ((), ())

    fn cluster_at(&self, _round_offset: std::num::NonZeroUsize) -> Vec<Self::Node> {


Rust is a great language with which to implement efficient and truly reliable, fault-tolerant services. And while there already are several Rust implementations of consensus algorithms, they are either rudimentary or incomplete or their API was not approachable enough for this author.


The project's priorities are as follows.

  1. Correctness

    The highest priority is correctness, which, in the context of consensus, requires stability, consistency and liveness.

    • Stability implies that once a node learns that a log entry a has been appended to the distributed log, it will never learn that a different entry b belongs in its place.
    • Consistency means that all nodes in the Paxakos network agree about the contents of their shared log. While nodes may temporarily fall behind, i.e. their log may be shorter than other nodes', they're logs must otherwise be equivalent.
    • Liveness means that the system won't get stuck, i.e. progress is always eventually made (assuming a there is no contention/some degree of cooperation).
  2. Simplicity

    Paxakos aims to be simple by providing few orthogonal primitives. To be clear, the goal is not to arbitrarily limit complexity. The goal is to have unentangled primitves; providing the same features with the least amount of complexity possible.

  3. Ergonomics

    Using Paxakos should be as pleasant as possible without sacrificing correctness or simplicity. The biggest challenge in this area are, at present, build times. If you have other concerns, please open an issue.


Paxakos is a Multi-Paxos implementation. It supports membership changes, concurrency, as well as taking and installing snapshots.

Membership Changes

The State trait exposes the cluster_at method. By implementing it, arbitrary membership changes may be made. No restrictions are imposed and it is up to users and implementors to make sure that any transition is safe.


Multi-Paxos allows for any number of rounds to be settled concurrently. This can be exploited by implementing State's concurrency method.

Please note that when concurrency is enabled, gaps may appear in the log. These must be closed before entries after them can be applied to the state. This is typically done by appending no-op entries. A utility for doing so automatically is provided, but its API is far from stable.


Consensus based clusters typically elect a single leader and who drive all progress. This is highly efficient, as each leader election incurs overhead. That notwithstanding, Paxakos has made the unusual design decision to allow multiple leaders at the same time, but for different rounds.

This design allows a follower node to "interject" an entry at the end of the concurrency window. This part of the design hasn't been fleshed out yet, but it could allow nodes to conveniently queue operations without introducing additional communication protocols besides Paxakos.


A node may take a snapshot of its current state. The result is a combination of the application specific State as well as pakakos specifc information. These snapshots may be used for graceful shutdown and restart, to bootstrap nodes which have joined the cluster or to catch up nodes that have fallen behind.


This section describes the Paxakos protocol. It is, for the most part, a typical Multi-Paxos protocol. Multi-Paxos generalizes Paxos to be run for multiple rounds, where each round represents a slot in the distributed log. Nodes may propose log entries to place in those slots. The liveness property guarantees that the cluster will, for each round, eventually converge on one of the proposed entries.

A central component of the protocol are coordination numbers. These are usually called "proposal numbers". However, because they are used throughout the protocol, Paxakos uses the more abstract term.

Appending an entry to the distributed log takes the following five steps.

  1. Prepare (RoundNum, CoordNum)

    In order for a node to append an entry to the distributed log, it must first become leader for the given round. If it already believes itself leader for the round, it will skip to step 3.

    To become leader for a round the node will send a prepare message containing the round number and a coordination number. The coordination number is chosen so that it is

    • higher than any previously encountered coordination number and
    • guaranteed not to be used by another node.

    The former is important for liveness. The latter is required for stability and consistency and is achieved by exploiting the deterministic order of nodes returned by cluster_at.

  2. Promise or Rejection

    When a node receives a prepare request, it checks that it hasn't accepted a previous such request with a coordination number that's equal or higher than the given one. If it hasn't, it sends back a promise not to accept any more proposals with a coordination number that's less the given one. Otherwise it sends back a rejection.

    1. Promise (Vec<(RoundNum, CoordNum, LogEntry)>)

      The promise is a set of triples, each consisting of a round number, a coordination number and a log entry. It can be thought to mean "I acknowledge your bid to become leader and give you my vote. However, in return you must respect these previous commitments I've made."

    2. Rejection (CoordNum, Option<LogEntry>)

      A rejection is sent with the highest observed coordination number so far. For the special case that the round has already converged and the node still has it available, it will send it along as well.

  3. Propose (RoundNum, CoordNum, LogEntry)

    When a node sent a prepare(r, c) request and received a quorum or more of promises in return (counting its own), it will believe itself to be leader for all rounds r... It may now start proposing log entries for any of these rounds, using c as the coordination number.

    The only restriction is that it must respect the promises it has received. If multiple promises contain a triple with the same round number, the one with the highest coordination number wins. (Triples with the same round and coordination number will have the same log entry as well.)

  4. Acceptance or Rejection

    When a node receives a proposal, it will check whether it has made any conflicting promises. If it hasn't it will simply reply that it has accepted the entry. Otherwise it will respond with a rejection.

    1. Acceptance ()

    2. Rejection (CoordNum, Option<LogEntry>)

      See 2.2.

  5. Commit (RoundNum, LogEntry::Id) / CommitById (RoundNum, LogEntry)

    After having received a quorum of acceptances, the round has converged on the proposed entry. The leader node will commit the entry locally and inform other nodes as well. Nodes who sent an acceptance will only be sent the log entry's id, others will receive the full entry.


The core algorithm of Paxakos is reasonably well-tested. However, "ancillary" features such as snapshots and passive mode are not well tested and presumably contain bugs. Also, APIs and serialized representations will likely change.

Use at your own risk.

Nightly Features

Paxakos currently relies on several nightly features. None of them are crucial, but there hasn't been any need become compatible with stable Rust.

Future Direction

Paxakos will probably remain dormant for the near future. This is because it needs to see some use and experimentation. Likeley exceptions are the following changes.

  • Improving Compile Times

    Any suggestions as to how comile times may be reduced are welcome. Compile times of dependent projects are the primary concern.

  • Adding comments and documentation

  • Adding convenience methods and decorations.

  • Support for Master Leases (see section 5.2 of Paxos Made Live).





Converged on log entry waiting to be applied.


A remote handle for a paxakos Node.


The default Node implementation.


A promise not to accept certain proposals anymore.


Used by Communicators to prepare replies.



A proposal could not be accepted.


Committing a log entry failed.


Emitted by Node's poll_events method.


A node's status, usually Leading or Following.


Preparing a round for proposals failed.


Rejection of a prepare request or a proposal.


Emitted by Shutdown's poll_shutdown method.



A coordination number.


Trait bound of log entry ids and node ids.


Appended to the shared log and applied to the shared State.


Builder to spawn a Node.


Describes a node in a Paxakos cluster.


Trait bound of both CoordNum as well as RoundNum.


A round number.


A Node that is being shut_down.


Distributed state to which log entries are applied.