Communication et configuration des nœuds

Erlang shines in building distributed systems! It allows multiple Erlang Virtual Machines (VMs), called nodes, to communicate seamlessly.

This lesson introduces how to set up and connect these nodes, forming the foundation for fault-tolerant and scalable applications.

Every Erlang node needs a unique name to identify itself in a distributed system. There are two types of names:

• Short names (e.g., mynode): Used for nodes on the same local machine or network segment. Started with -sname.
• Long names (e.g., node1@example.com): Essential for nodes distributed across different hosts and networks. Requires a fully qualified domain name (FQDN) and started with -name.

We'll primarily use short names for local examples.

For two Erlang nodes to communicate, they must share a common secret key called the Erlang cookie. Think of it like a password for node-to-node authentication.

• If nodes have different cookies, they cannot connect.
• The cookie is a string (e.g., mysecretcookie).
• It's crucial for security: keep your cookies secret and generate strong ones!

Let's start an Erlang node with a short name and a cookie. Open your terminal and run:

Once the Erlang shell loads, you can check your node's name and cookie:

• node(). (shows the node's full name)
• erlang:get_cookie(). (shows the cookie)

When an Erlang node starts, it registers itself with a local process called EPMD (Erlang Port Mapper Daemon).

EPMD's job is to keep track of which Erlang nodes are running on the local host and which TCP ports they are listening on. When one node wants to connect to another on the same machine, it asks EPMD for the target node's port number.

You can see registered nodes with epmd -names in a separate terminal.

Now, start a second node in another terminal. Make sure to use the same cookie:

From the alpha node's shell, try to ping beta:

Replace ~s with your machine's hostname (e.g., net_adm:ping(beta@localhost).). If it returns pong, they are connected! You can also use nodes(). to see connected nodes.

This simple Erlang module helps identify the current node. It's a full program that reports its own node name.

Save this as node_id.erl, compile it in your Erlang shell (c(node_id).), and then call node_id:print_name(). to see the output.

For connecting nodes across different physical machines or networks, you must use long node names with the -name flag. This requires using the node's fully qualified domain name (FQDN) or IP address.

Example: erl -name node1@server.example.com -setcookie myappsecret

Using long names ensures that nodes can be uniquely identified and reached over a wider network.

While the Erlang cookie provides basic authentication, consider these points:

• Strong Cookies: Use long, random strings for production environments.
• Firewall Rules: Configure firewalls to allow EPMD (port 4369) and the dynamic Erlang node ports to communicate only between trusted hosts.
• TLS: For sensitive data or untrusted networks, use Transport Layer Security (TLS) for encrypted communication (covered in a later lesson).

You've learned about Erlang nodes, naming, cookies, and EPMD. Let's test your understanding!

You've taken the first step into distributed Erlang! We covered:

• Erlang Nodes: Separate Erlang VMs that can communicate.
• Node Names: -sname for local, -name for remote (FQDN).
• Erlang Cookie: A shared secret for node authentication.
• EPMD: The daemon for local node discovery.
• Connecting Nodes: Using net_adm:ping/1 to establish communication.

Next, we'll explore how these connected nodes can actually talk to each other!