Switch branches/tags
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Tutorial - Understanding Circuit Relay

Welcome! This tutorial will help you understand circuit relay, where it fits in the stack and how to use it.

So what is a circuit-relay and what do we need it for?

In p2p networks there are many cases where two nodes can't talk to each other directly. That may happen because of network topology, i.e. NATs, or execution environments - for example browser nodes can't connect to each other directly because they lack any sort of socket functionality and relaying on specialized rendezvous nodes introduces an undesirable centralization point to the network. A circuit-relay is a way to solve this problem - it is a node that allows two other nodes that can't otherwise talk to each other, use a third node, a relay to do so.

How does circuit relay work?

for a more in-depth explanation take a look at the relay spec and js-libp2p-circuit README

Here is a simple diagram depicting how a typical circuit-relay connection might look:

+---------------------+         |          |         +---------------------+
|       Node A        |---------> FIREWALL <---------|        Node B       |
+----------^----------+         |          |         +----------^----------+
           |                                                    |           
           |               +---------------------+              |           
           +--------------->   Circuit Relay     <--------------+           

Node A tries to connect to Node B but, UH-OH! There is a firewall in between that's preventing it from happening. If both Node A and Node B know about a relay, they can use it to establish the connection.

This is what it looks like, in simplified steps:

  1. Node A tries to connect to Node B over one of its known addresses
  2. Connection fails because of firewall/NAT/incompatible transports/etc...
  3. Both Node A and Node B know of the same relay - Relay
  4. Node A falls back to dialing over Relay to Node B using its '/p2p-circuit' address, which involves:
    1. Node A sends a HOP request to Relay
    2. Relay extracts the destination address, figures out that a circuit to Node B is being requested
    3. Relay sends a STOP request to Node B
    4. Node B responds with a SUCCESS message
    5. Relay proceed to create a circuit over the two nodes
  5. Node A and Node B are now connected over Relay

That's it!

What's up with this HOP and STOP?

Circuit relay consists of two logical parts — dialer/listener and relay (HOP). The listener is also known as the STOP node. Each of these — dial, listen, and relay — happen on a different node. If we use the nodes from the above example, it looks something like this:

  • The dialer knows how to dial a relay (HOP) - Node A
  • The relay (HOP) knows how to contact a destination node (STOP) and create a circuit - Relay node
  • The listener (STOP) knows how to process relay requests that come from the relay (HOP) node - Node B

Fun fact - the HOP and STOP names are also used internally by circuit to identify the network message types.

A few caveats (and features)

There are a couple of caveats and features to be aware of:

  • A Relay will only work if it already has a connection to the STOP node
  • No multihop dialing is supported. It's a feature planed for upcoming releases (no date on this one)
    • multihop dialing is when several relays are used to establish the connection
  • It is possible to use explicit relay addresses to connect to a node, or even to listen for connections on. See next section to learn how to do this.

A word on circuit relay addresses

A circuit relay address is a multiaddress that describes how to either connect to a peer over a relay (or relays), or allow a peer to announce it is reachable over a particular relay or any relay it is already connected to.

Circuit relay addresses are very flexible and can describe many different aspects of how to esablish the relayed connection. In its simplest form, it looks something like this:

  • /p2p-circuit/ipfs/QmPeer

If we want to be specific as to which transport we want to use to establish the relay, we can encode that in the address as well:

  • /ip4/

This tells us that we want to use QmRelay located at address and port 65000.

  • /ip4/

We can take it a step further and encode the same information for the destination peer. In this case, we have it located at on port 8080 and using a Web sockets transport!

  • /ip4/

If a node is configured with this address, it will use the specified host (/ip4/ as a relay and it will be reachable over this relay.

  • There could multiple addresses of this sort specified in the config, in which case the node will be reachable over all of them.
  • This is useful if, for example, the node is behind a firewall but wants to be reachable from the outside over a specific relay.

Other use-cases are also supported by this scheme, e.g. we can have multiple hops (circuit-relay nodes) encoded in the address, something planed for future releases.

Step-by-step instructions

Here's what we are going to be doing, today:

  1. Install and configure go-ipfs and js-ipfs nodes
  2. Configure and run the js or go ipfs node
  3. Configure and run the bundled example
  4. Connect the two browser nodes to the circuit relay
  5. Dial the two browser nodes using a /p2p-circuit address
  6. Finally, send data from one browser using the bundled example!

We should end up with something similar to the bellow screenshot after we've gone through all the steps:

Let's go.

1. Set up

You'll need to have an implementation of IPFS running on your machine. Currently, this means either go-ipfs or js-ipfs.

Installing go-ipfs can be done by installing the binary here. Alternatively, you could follow the instructions in the README at ipfs/go-ipfs.

Installing js-ipfs requires you to have node and npm. Then, you simply run:

> npm install --global ipfs
> jsipfs --help

This will alias jsipfs on your machine; this is to avoid issues with go-ipfs being called ipfs.

At this point, you have either js-ipfs or go-ipfs running. Now, initialize it:

> ipfs init
# or
> jsipfs init

This will set up your IPFS repo in your home directory.

Configure and run the js or go ipfs node

You can use a go-ipfs or a js-ipfs node as a relay. We'll demonstrate how to set both up in this tutorial and we encourage you to try both out. That said, either js or go should do the trick!

Setting up a go-ipfs node

In order to enable the relay functionality in go-ipfs we need to edit it's configuration file, located under ~/.ipfs/config:

  "Swarm": {
    "AddrFilters": null,
    "ConnMgr": {
      "GracePeriod": "20s",
      "HighWater": 900,
      "LowWater": 600,
      "Type": "basic"
    "DisableBandwidthMetrics": false,
    "DisableNatPortMap": false,
    "DisableRelay": false,
    "EnableRelayHop": true

The two options we're looking for are DisableRelay and EnableRelayHop. We want the former (DisableRelay) set to false and the latter (EnableRelayHop) to true, just like in the example above. That should set our go node as a relay.

We also need to make sure our go node can be dialed from the browser. For that, we need to enable a transport that both the browser and the go node can communicate over. We will use the web sockets transport, although there are others that can be used, such as webrtc-star and websocket-star. To enable the transport and set the interface and port we need to edit the ~/.ipfs/config one more time. Let's find the Swarm array and add our desired address there. I picked /ip4/ because it is a port I know is not being used by anything on my machine, but we can also use port 0 so that the OS chooses a random available port for us — either one should work.

  "Swarm": [

The config should look similar to the above snippet after we've edited it.

Setting up a js-ipfs node

We need to go through similar steps to enable circuit relay in jsipfs. However, the config options are slightly different — that should change once this feature is not marked as experimental, but for now we have to deal with two different sets of options.

Just as we did with go-ipfs, go ahead and edit js-ipfs config file located under ~/.jsipfs/config. Let's add the following config:

  "relay": {
    "enabled": true,
    "hop": {
      "enabled": true

Note that we don't have to do anything to enable the websocket transport as it is enabled by default in jsipfs.

Starting the relay node

We can start the relay nodes by either running ipfs daemon or jsipfs daemon:

go ipfs

$ ipfs daemon
Initializing daemon...
Swarm listening on /ip4/
Swarm listening on /ip4/
Swarm listening on /ip6/::1/tcp/4001
Swarm listening on /p2p-circuit/ipfs/QmY73BLYav2gYc9PCEnjQqbfSGiqFv3aMsRXNyKFGtUoGF
Swarm announcing /ip4/
Swarm announcing /ip4/
Swarm announcing /ip4/
Swarm announcing /ip6/::1/tcp/4001
API server listening on /ip4/
Gateway (readonly) server listening on /ip4/
Daemon is ready

In the case of go ipfs, the crucial /ipfs/Qm... part of the multiaddr might be missing. In that case, you can get it by running the ipfs id command.

$ ipfs id
        "ID": "QmY73BLYav2gYc9PCEnjQqbfSGiqFv3aMsRXNyKFGtUoGF",
        "PublicKey": "CAASpgIwggEiMA0GCSqGSIb3DQEBAQUAA4IBDwAwggEKAoIBAQC84qPFzqajCfnvaJunqt48S1LIBRthXV60q5QClL+dUfOOU/m7v1ZcpNhvFFUN6tVCDaoT5AxEv0czxZiVx/njl6FVIc6tE1J+HWpc8cbAXNY6QbbyzKl/rjp7V8/QClE0JqgjIk84wnWGTwFhOEt0hnpu2XFt9iHaenSfg3EAa8K9MlbxmbawuxNLJJf7VZXkJrUNl6WOglAVU8Sqc4QaahCLVK5Dzo98zDBq1KDBxMbUgH0LTqzr6i+saxkEHZmBKO+mMVT3LzOUx1DQR4pLAw1qgoJstsIZEaJ2XLh975IiI7OKqWYH7+3NyNK2sldJK/4Zko4rH3irmnkAxLcFAgMBAAE=",
        "Addresses": [
        "AgentVersion": "go-ipfs/0.4.14-dev/cb5bb7dd8",
        "ProtocolVersion": "ipfs/0.1.0"

We can then grab the resolved multiaddr from the Addresses array — /ip4/ Let's note it down somewhere and move to the next step.

js ipfs

$ jsipfs daemon
Initializing daemon...
Swarm listening on /p2p-circuit/ipfs/QmfQj8YwDdy1uP2DpZBa7k38rSGPvhHiC52cdAGWBqoVpq
Swarm listening on /p2p-circuit/ip4/
Swarm listening on /p2p-circuit/ip4/
Swarm listening on /ip4/
Swarm listening on /ip4/
Swarm listening on /ip4/
API is listening on: /ip4/
Gateway (readonly) is listening on: /ip4/
Daemon is ready

Look out for an address similar to /ip4/ Note it down somewhere, and let's move on to the next step.

2. Configure and run the bundled example

Now that we have ipfs installed and initialized, let's set up the included example. This is a standard npm package, so the usual npm install should get us going. Let's cd into the examples/circuit-relaying directory and run:

npm install

After it finishes, we should be able to run the project with npm start and get output similar to:

npm run start
Server running at http://localhost:1234

The bundled example is a simple chat app that uses another cool ipfs feature - pubsub. Let's open up a browser and paste the above url into the address bar. We should see something similar to the following image:

3. Connect the two browser nodes to the circuit relay

In order for our browser nodes to be able to messages each other, we need to get them connected. But to do that, we need to use a relay - browser nodes can't be connected directly because of lack of socket support.

Remember the caveat above Currently a Relay will only work if it already has a connection to the STOP node? This means that we need to connect our browser nodes to the relay node first.

Having both browsers running side by side (as shown in the first screenshot), enter the /ip4/ address noted above into the Connect to Peer field and hit the Connect button:

After connecting to the IPFS node, we should see the relay peer show up under the Peers Connected box.

Let's repeat the same steps with the second tab. Now, both of our browser nodes should be connected to the relay and we can move on to the next step.

4. Dial the two browser nodes using a /p2p-circuit address

Now that our browsers are both connected to the relay peer, let's get them connected to each other. Head out to the Addresses box in one of the tabs, copy the /p2p-circuit address and then paste it into the Connect to Peer box in the other tab. Repeat these steps on the second tab.

Let's hit the Connect button on each of the tabs and we should get the two browsers connected and join the chat room.

5. Send data browser to browser.

Now that we have the two browsers connected, let's try the app out. Type a few words in one of the browser windows and you should see them appear in the other as well!

Thats it!

So what just happened?

Good question!

  • We used js-ipfs running in the browser with circuit relay enabled:
    • Notice the relay.enabled below

you can find it in src/app.js

const ipfs = new IPFS({
  repo: repo(),
  relay: {
    enabled: true,
    hop: {
      enabled: true
  config: {
    Bootstrap: []
  • We connected the browser nodes to an external node over its websocket transport using the /ip4/ multiaddr. That external node happens to be a HOP node, meaning that it can relay connections for our browsers (and other nodes) allowing them to connect

  • And finally we connected the two browser nodes using the /p2p-circuit/ipfs/... multiaddr. Take a look at the code below in src/app.js - lines 102-107

ipfs.swarm.connect(peer, (err) => {
  if (err) {
    return console.error(err)
  $pAddrs.innerHTML += `<li>${peer.trim()}</li>`

The above code snippet handles connecting to other nodes using ipfs.swarm.connect. Notice how there wasn't anything special we had to do to use the circuit once we had everything connected; all the magic is in the multiaddr! Multiaddrs are AWESOME!

I encourage the reader to take a look at the bundled app code to see how the browser nodes get setup, suffice to say nothing changes from the perspective of using an IPFS node in js code, apart from the new EXPERIMENTAL options.

Finally, a side note on pubsub. We've used the amazing ipfs-pubsub-room module, to enable the chat functionality. Make sure to take a look at the demo video that explains how pubsub works and how it can be used to build other applications!