matrix-docker-ansible-deploy/docs/configuring-well-known.md
kristbaum 201292cdc5
typo
2020-06-07 19:56:12 +02:00

8.5 KiB

Configuring Service Discovery via .well-known

Service discovery is a way for the Matrix network to discover where a Matrix server is.

There are 2 types of well-known service discovery that Matrix makes use of:

  • (important) Federation Server discovery (/.well-known/matrix/server) -- assists other servers in the Matrix network with finding your server. Without a proper configuration, your server will effectively not be part of the Matrix network. Learn more in Introduction to Federation Server Discovery

  • (not that important) Client Server discovery (/.well-known/matrix/client) -- assists programs that you use to connect to your server (e.g. Riot), so that they can make it more convenient for you by automatically configuring the "Homeserver URL" and "Identity Server URL" addresses. Learn more in Introduction to Client Server Discovery

Introduction to Federation Server Discovery

All services created by this playbook are meant to be installed on their own server (such as matrix.<your-domain>).

As per the Server-Server specification, to use a Matrix user identifier like @<username>:<your-domain> while hosting services on a subdomain like matrix.<your-domain>, the Matrix network needs to be instructed of such delegation/redirection.

Server delegation can be configured using DNS SRV records or by setting up a /.well-known/matrix/server file on the base domain (<your-domain.com>).

Both methods have their place and will continue to do so. You only need to use just one of these delegation methods. For simplicity reasons, our setup advocates for the /.well-known/matrix/server method and guides you into using that.

To learn how to set up /.well-known/matrix/server, read the Installing section below.

Introduction to Client Server Discovery

Client Server Service discovery lets various client programs which support it, to receive a full user id (e.g. @username:example.com) and determine where the Matrix server is automatically (e.g. https://matrix.example.com).

This lets you (and your users) easily connect to your Matrix server without having to customize connection URLs. When using client programs that support it, you won't need to point them to https://matrix.example.com in Custom Server options manually anymore. The connection URL would be discovered automatically from your full username.

As per the Client-Server specification Matrix does Client Server service discovery using a /.well-known/matrix/client file hosted on the base domain (e.g. example.com).

However, this playbook installs your Matrix server on another domain (e.g. matrix.example.com) and not on the base domain (e.g. example.com), so it takes a little extra manual effort to set up the file.

To learn how to set it up, read the Installing section below.

Installing well-known files on the base domain's server

To implement the two service discovery mechanisms, your base domain's server (e.g. example.com) needs to run an HTTPS-capable webserver.

If you don't have a server for your base domain at all, you can use the Matrix server for this. See Serving the base domain to learn how the playbook can help you set it up. If you decide to go this route, you don't need to read ahead in this document. When Serving the base domain, the playbook takes care to serve the appropriate well-known files automatically.

If you're managing the base domain by yourself somehow, you'll need to set up serving of some /.well-known/matrix/* files from it via HTTPS.

To make things easy for you to set up, this playbook generates and hosts 2 well-known files on the Matrix domain's server (e.g. https://matrix.example.com/.well-known/matrix/server and https://matrix.example.com/.well-known/matrix/client), even though this is the wrong place to host them.

You have 2 options when it comes to installing the files on the base domain's server:

(Option 1): Copying the files manually to your base domain's server

Hint: Option 2 (below) is generally a better way to do this. Make sure to go with that one, if possible.

All you need to do is:

  • copy /.well-known/matrix/server and /.well-known/matrix/client from the Matrix server (e.g. matrix.example.com) to your base domain's server (example.com). You can find these files in the /matrix/static-files/.well-known/matrix directory on the Matrix server. They are also accessible on URLs like this: https://matrix.example.com/.well-known/matrix/server (same for client).

  • set up the server at your base domain (e.g. example.com) so that it adds an extra HTTP header when serving the /.well-known/matrix/client file. CORS, the Access-Control-Allow-Origin header should be set with a value of *. If you don't do this step, web-based Matrix clients (like Riot) may fail to work. Setting up headers for the /.well-known/matrix/server file is not necessary, as this file is only consumed by non-browsers, which don't care about CORS.

This is relatively easy to do and possibly your only choice if you can only host static files from the base domain's server. It is, however, a little fragile, as future updates performed by this playbook may regenerate the well-known files and you may need to notice that and copy them over again.

(Option 2): Setting up reverse-proxying of the well-known files from the base domain's server to the Matrix server

This option is less fragile and generally better.

On the base domain's server (e.g. example.com), you can set up reverse-proxying, so that any access for the /.well-known/matrix location prefix is forwarded to the Matrix domain's server (e.g. matrix.example.com).

With this method, you don't need to add special HTTP headers for CORS reasons (like Access-Control-Allow-Origin), because your Matrix server (where requests ultimately go) will be configured by this playbook correctly.

For nginx, it would be something like this:

# This is your HTTPS-enabled server for DOMAIN.
server {
	server_name DOMAIN;

	location /.well-known/matrix {
		proxy_pass https://matrix.DOMAIN/.well-known/matrix;
		proxy_set_header X-Forwarded-For $remote_addr;
	}

	# other configuration
}

For Apache, it would be something like this:

<VirtualHost *:443>
	ServerName DOMAIN

	SSLProxyEngine on
	<Location /.well-known/matrix>
		ProxyPass "https://matrix.DOMAIN/.well-known/matrix"
	</Location>

	# other configuration
</VirtualHost>

For Caddy 2, it would be something like this:

reverse_proxy /.well-known/matrix/* https://matrix.DOMAIN {
	header_up Host {http.reverse_proxy.upstream.hostport}
}

For Caddy 1, it would be something like this:

proxy /.well-known/matrix/ https://matrix.DOMAIN {
    header_upstream Host {http.reverse_proxy.upstream.hostport}
}

For HAProxy, it would be something like this:

frontend www-https
	# Select a Challenge for Matrix federation redirect
	acl matrix-acl path_beg /.well-known/matrix/
	# Use the challenge backend if the challenge is set
	use_backend matrix-backend if matrix-acl
backend matrix-backend
	# Redirects the .well-known matrix to the matrix server for federation.
	http-request set-header Host matrix.example.com
	server matrix matrix.example.com:80
	# Map url path as ProxyPass does
	reqirep ^(GET|POST|HEAD)\ /.well-known/matrix/(.*) \1\ /\2
	# Rewrite redirects as ProxyPassReverse does
	acl response-is-redirect res.hdr(Location) -m found
	rsprep ^Location:\ (http|https)://matrix.example.com\/(.*) Location:\ \1://matrix.example.com/.well-known/matrix/\2 if response-is-redirect

Make sure to:

  • replace DOMAIN in the server configuration with your actual domain name
  • and: to do this for the HTTPS-enabled server block, as that's where Matrix expects the file to be

Confirming it works

No matter which method you've used to set up the well-known files, if you've done it correctly you should be able to see a JSON file at both of these URLs:

  • https://<domain>/.well-known/matrix/server
  • https://<domain>/.well-known/matrix/client

You can also check if everything is configured correctly, by checking if services work.