Module libp2p::tutorials::ping

source ·
Expand description

Ping Tutorial - Getting started with rust-libp2p

This tutorial aims to give newcomers a hands-on overview of how to use the Rust libp2p implementation. People new to Rust likely want to get started on Rust itself, before diving into all the networking fun. This library makes heavy use of asynchronous Rust. In case you are not familiar with this concept, the Rust async-book should prove useful. People new to libp2p might prefer to get a general overview at libp2p.io first, although libp2p knowledge is not required for this tutorial.

We are going to build a small ping clone, sending a ping to a peer, expecting a pong as a response.

Scaffolding

Let’s start off by

  1. Updating to the latest Rust toolchain, e.g.: rustup update

  2. Creating a new crate: cargo init rust-libp2p-tutorial

  3. Adding libp2p as well as futures as dependencies in the Cargo.toml file. Current crate versions may be found at crates.io. We will also include async-std with the “attributes” feature to allow for an async main. At the time of writing we have:

    [package]
    name = "rust-libp2p-tutorial"
    version = "0.1.0"
    edition = "2021"

[dependencies]
    libp2p = { version = "0.52", features = ["tcp", "tls", "dns", "async-std", "noise", "yamux", "websocket", "ping", "macros"] }
    futures = "0.3.21"
    async-std = { version = "1.12.0", features = ["attributes"] }
    tracing-subscriber = { version = "0.3", features = ["env-filter"] }

Network identity

With all the scaffolding in place, we can dive into the libp2p specifics. First we need to create a network identity for our local node in async fn main(), annotated with an attribute to allow main to be async. Identities in libp2p are handled via a public/private key pair. Nodes identify each other via their PeerId which is derived from their public key. Now, replace the contents of main.rs by:

use std::error::Error;
use tracing_subscriber::EnvFilter;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity();

    Ok(())
}

Go ahead and build and run the above code with: cargo run. Nothing happening thus far.

Transport

Next up we need to construct a transport. Each transport in libp2p provides encrypted streams. E.g. combining TCP to establish connections, TLS to encrypt these connections and Yamux to run one or more streams on a connection. Another libp2p transport is QUIC, providing encrypted streams out-of-the-box. We will stick to TCP for now. Each of these implement the Transport trait.

use std::error::Error;
use tracing_subscriber::EnvFilter;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
        .with_async_std()
        .with_tcp(
            libp2p::tcp::Config::default(),
            libp2p::tls::Config::new,
            libp2p::yamux::Config::default,
        )?;

    Ok(())
}

Network behaviour

Now it is time to look at another core trait of rust-libp2p: the NetworkBehaviour. While the previously introduced trait Transport defines how to send bytes on the network, a NetworkBehaviour defines what bytes and to whom to send on the network.

To make this more concrete, let’s take a look at a simple implementation of the NetworkBehaviour trait: the ping::Behaviour. As you might have guessed, similar to the good old ICMP ping network tool, libp2p ping::Behaviour sends a ping to a peer and expects to receive a pong in turn. The ping::Behaviour does not care how the ping and pong messages are sent on the network, whether they are sent via TCP, whether they are encrypted via noise or just in plaintext. It only cares about what messages and to whom to sent on the network.

The two traits Transport and NetworkBehaviour allow us to cleanly separate how to send bytes from what bytes and to whom to send.

With the above in mind, let’s extend our example, creating a ping::Behaviour at the end:

use libp2p::ping;
use tracing_subscriber::EnvFilter;
use std::error::Error;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
        .with_async_std()
        .with_tcp(
            libp2p::tcp::Config::default(),
            libp2p::tls::Config::new,
            libp2p::yamux::Config::default,
        )?
        .with_behaviour(|_| ping::Behaviour::default())?;

    Ok(())
}

Swarm

Now that we have a Transport and a NetworkBehaviour, we can build the Swarm which connects the two, allowing both to make progress. Put simply, a Swarm drives both a Transport and a NetworkBehaviour forward, passing commands from the NetworkBehaviour to the Transport as well as events from the Transport to the NetworkBehaviour.

use libp2p::ping;
use std::error::Error;
use tracing_subscriber::EnvFilter;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
        .with_async_std()
        .with_tcp(
            libp2p::tcp::Config::default(),
            libp2p::tls::Config::new,
            libp2p::yamux::Config::default,
        )?
        .with_behaviour(|_| ping::Behaviour::default())?
        .build();

    Ok(())
}

Idle connection timeout

Now, for this example in particular, we need set the idle connection timeout. Otherwise, the connection will be closed immediately.

Whether you need to set this in your application too depends on your usecase. Typically, connections are kept alive if they are “in use” by a certain protocol. The ping protocol however is only an “auxiliary” kind of protocol. Thus, without any other behaviour in place, we would not be able to observe the pings.

use libp2p::ping;
use std::error::Error;
use std::time::Duration;
use tracing_subscriber::EnvFilter;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
        .with_async_std()
        .with_tcp(
            libp2p::tcp::Config::default(),
            libp2p::tls::Config::new,
            libp2p::yamux::Config::default,
        )?
        .with_behaviour(|_| ping::Behaviour::default())?
        .with_swarm_config(|cfg| cfg.with_idle_connection_timeout(Duration::from_secs(30))) // Allows us to observe pings for 30 seconds.
        .build();

    Ok(())
}

Multiaddr

With the Swarm in place, we are all set to listen for incoming connections. We only need to pass an address to the Swarm, just like for std::net::TcpListener::bind. But instead of passing an IP address, we pass a Multiaddr which is yet another core concept of libp2p worth taking a look at.

A Multiaddr is a self-describing network address and protocol stack that is used to establish connections to peers. A good introduction to Multiaddr can be found at docs.libp2p.io/concepts/addressing and its specification repository github.com/multiformats/multiaddr.

Let’s make our local node listen on a new socket. This socket is listening on multiple network interfaces at the same time. For each network interface, a new listening address is created. These may change over time as interfaces become available or unavailable. For example, in case of our TCP transport it may (among others) listen on the loopback interface (localhost) /ip4/127.0.0.1/tcp/24915 as well as the local network /ip4/192.168.178.25/tcp/24915.

In addition, if provided on the CLI, let’s instruct our local node to dial a remote peer.

use libp2p::{ping, Multiaddr};
use std::error::Error;
use std::time::Duration;
use tracing_subscriber::EnvFilter;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
        .with_async_std()
        .with_tcp(
            libp2p::tcp::Config::default(),
            libp2p::tls::Config::new,
            libp2p::yamux::Config::default,
        )?
        .with_behaviour(|_| ping::Behaviour::default())?
        .with_swarm_config(|cfg| cfg.with_idle_connection_timeout(Duration::from_secs(30))) // Allows us to observe pings for 30 seconds.
        .build();

    // Tell the swarm to listen on all interfaces and a random, OS-assigned
    // port.
    swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;

    // Dial the peer identified by the multi-address given as the second
    // command-line argument, if any.
    if let Some(addr) = std::env::args().nth(1) {
        let remote: Multiaddr = addr.parse()?;
        swarm.dial(remote)?;
        println!("Dialed {addr}")
    }

    Ok(())
}

Continuously polling the Swarm

We have everything in place now. The last step is to drive the Swarm in a loop, allowing it to listen for incoming connections and establish an outgoing connection in case we specify an address on the CLI.

use futures::prelude::*;
use libp2p::swarm::SwarmEvent;
use libp2p::{ping, Multiaddr};
use std::error::Error;
use std::time::Duration;
use tracing_subscriber::EnvFilter;

#[async_std::main]
async fn main() -> Result<(), Box<dyn Error>> {
    tracing_subscriber::fmt().with_env_filter(EnvFilter::from_default_env()).init();

    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
        .with_async_std()
        .with_tcp(
            libp2p::tcp::Config::default(),
            libp2p::tls::Config::new,
            libp2p::yamux::Config::default,
        )?
        .with_behaviour(|_| ping::Behaviour::default())?
        .with_swarm_config(|cfg| cfg.with_idle_connection_timeout(Duration::from_secs(30))) // Allows us to observe pings for 30 seconds.
        .build();

    // Tell the swarm to listen on all interfaces and a random, OS-assigned
    // port.
    swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;

    // Dial the peer identified by the multi-address given as the second
    // command-line argument, if any.
    if let Some(addr) = std::env::args().nth(1) {
        let remote: Multiaddr = addr.parse()?;
        swarm.dial(remote)?;
        println!("Dialed {addr}")
    }

    loop {
        match swarm.select_next_some().await {
            SwarmEvent::NewListenAddr { address, .. } => println!("Listening on {address:?}"),
            SwarmEvent::Behaviour(event) => println!("{event:?}"),
            _ => {}
        }
    }
}

Running two nodes

For convenience the example created above is also implemented in full in examples/ping.rs. Thus, you can either run the commands below from your own project created during the tutorial, or from the root of the rust-libp2p repository. Note that in the former case you need to ignore the --example ping argument.

You need two terminals. In the first terminal window run:

cargo run --example ping

It will print the new listening addresses, e.g.

Listening on "/ip4/127.0.0.1/tcp/24915"
Listening on "/ip4/192.168.178.25/tcp/24915"
Listening on "/ip4/172.17.0.1/tcp/24915"
Listening on "/ip6/::1/tcp/24915"

In the second terminal window, start a new instance of the example with:

cargo run --example ping -- /ip4/127.0.0.1/tcp/24915

Note: The Multiaddr at the end being one of the Multiaddr printed earlier in terminal window one. Both peers have to be in the same network with which the address is associated. In our case any printed addresses can be used, as both peers run on the same device.

The two nodes will establish a connection and send each other ping and pong messages every 15 seconds.