net.Server

class net.Server extends EventEmitter

This class is used to create a TCP or IPC server.

A listening TCP net.Server can be transferred to a worker thread by listing it in the transferList of a worker_threads postMessage() call. This moves the underlying listening socket to the receiving thread, where it resumes accepting connections. See Transferring TCP handles to other threads.

new net.Server(options?, connectionListener?): net.Server

net.Server is an EventEmitter with the following events:

Emitted when the server closes. If connections exist, this event is not emitted until all connections are ended.

Emitted when a new connection is made. socket is an instance of net.Socket.

Emitted when an error occurs. Unlike net.Socket, the 'close' event will not be emitted directly following this event unless server.close() is manually called. See the example in discussion of server.listen().

Emitted when the server has been bound after calling server.listen().

When the number of connections reaches the threshold of server.maxConnections, the server will drop new connections and emit 'drop' event instead. If it is a TCP server, the argument is as follows, otherwise the argument is undefined.

server.address(): Object | string | null

Returns the bound address, the address family name, and port of the server as reported by the operating system if listening on an IP socket (useful to find which port was assigned when getting an OS-assigned address): { port: 12346, family: 'IPv4', address: '127.0.0.1' }.

For a server listening on a pipe or Unix domain socket, the name is returned as a string.

const server = net.createServer((socket) => {
  socket.end('goodbye\n');
}).on('error', (err) => {
  // Handle errors here.
  throw err;
});

// Grab an arbitrary unused port.
server.listen(() => {
  console.log('opened server on', server.address());
});

server.address() returns null before the 'listening' event has been emitted or after calling server.close().

server.close(callback?): net.Server

Stops the server from accepting new connections and keeps existing connections. This function is asynchronous, the server is finally closed when all connections are ended and the server emits a 'close' event. The optional callback will be called once the 'close' event occurs. Unlike that event, it will be called with an Error as its only argument if the server was not open when it was closed.

server[Symbol.asyncDispose](): void

Calls server.close() and returns a promise that fulfills when the server has closed.

server.getConnections(callback): net.Server

Asynchronously get the number of concurrent connections on the server. Works when sockets were sent to forks.

Callback should take two arguments err and count.

server.listen(): void

Start a server listening for connections. A net.Server can be a TCP or an IPC server depending on what it listens to.

Possible signatures:

• server.listen(handle[, backlog][, callback])
• server.listen(options[, callback])
• server.listen(path[, backlog][, callback]) for IPC servers
• server.listen([port[, host[, backlog]]][, callback]) for TCP servers

This function is asynchronous. When the server starts listening, the 'listening' event will be emitted. The last parameter callback will be added as a listener for the 'listening' event.

All listen() methods can take a backlog parameter to specify the maximum length of the queue of pending connections. The actual length will be determined by the OS through sysctl settings such as tcp_max_syn_backlog and somaxconn on Linux. The default value of this parameter is 511 (not 512).

All net.Socket are set to SO_REUSEADDR (see socket(7) for details).

The server.listen() method can be called again if and only if there was an error during the first server.listen() call or server.close() has been called. Otherwise, an ERR_SERVER_ALREADY_LISTEN error will be thrown.

One of the most common errors raised when listening is EADDRINUSE. This happens when another server is already listening on the requested port/path/handle. One way to handle this would be to retry after a certain amount of time:

server.on('error', (e) => {
  if (e.code === 'EADDRINUSE') {
    console.error('Address in use, retrying...');
    setTimeout(() => {
      server.close();
      server.listen(PORT, HOST);
    }, 1000);
  }
});

server.listen(handle, backlog?, callback?): net.Server

Start a server listening for connections on a given handle that has already been bound to a port, a Unix domain socket, or a Windows named pipe.

The handle object can be either a server, a socket (anything with an underlying _handle member), a BoundSocket, or an object with an fd member that is a valid file descriptor.

When handle is a BoundSocket, the server adopts the already-bound socket and starts listening on it. Adoption consumes the bound socket (see ownership transfer).

Listening on a file descriptor is not supported on Windows.

server.listen(options, callback?): net.Server

If handle is specified, the server adopts that pre-bound socket. Otherwise, if port is specified, it behaves the same as server.listen([port[, host[, backlog]]][, callback]). Otherwise, if path is specified, it behaves the same as server.listen(path[, backlog][, callback]). If none of them is specified, an error will be thrown.

If exclusive is false (default), then cluster workers will use the same underlying handle, allowing connection handling duties to be shared. When exclusive is true, the handle is not shared, and attempted port sharing results in an error. An example which listens on an exclusive port is shown below.

server.listen({
  host: 'localhost',
  port: 80,
  exclusive: true,
});

When exclusive is true and the underlying handle is shared, it is possible that several workers query a handle with different backlogs. In this case, the first backlog passed to the master process will be used.

Starting an IPC server as root may cause the server path to be inaccessible for unprivileged users. Using readableAll and writableAll will make the server accessible for all users.

If the signal option is enabled, calling .abort() on the corresponding AbortController is similar to calling .close() on the server:

const controller = new AbortController();
server.listen({
  host: 'localhost',
  port: 80,
  signal: controller.signal,
});
// Later, when you want to close the server.
controller.abort();

server.listen(path, backlog?, callback?): net.Server

Start an IPC server listening for connections on the given path.

server.listen(port?, host?, backlog?, callback?): net.Server

Start a TCP server listening for connections on the given port and host.

If port is omitted or is 0, the operating system will assign an arbitrary unused port, which can be retrieved by using server.address().port after the 'listening' event has been emitted.

If host is omitted, the server will accept connections on the unspecified IPv6 address (::) when IPv6 is available, or the unspecified IPv4 address (0.0.0.0) otherwise.

In most operating systems, listening to the unspecified IPv6 address (::) may cause the net.Server to also listen on the unspecified IPv4 address (0.0.0.0).

When the number of connections reaches the server.maxConnections threshold:

1. If the process is not running in cluster mode, Node.js will close the connection.

2. If the process is running in cluster mode, Node.js will, by default, route the connection to another worker process. To close the connection instead, set server.dropMaxConnection to true.

It is not recommended to use this option once a socket has been sent to a child with child_process.fork().

Set this property to true to begin closing connections once the number of connections reaches the server.maxConnections threshold. This setting is only effective in cluster mode.

server.ref(): net.Server

Opposite of unref(), calling ref() on a previously unrefed server will not let the program exit if it's the only server left (the default behavior). If the server is refed calling ref() again will have no effect.

server.unref(): net.Server

Calling unref() on a server will allow the program to exit if this is the only active server in the event system. If the server is already unrefed calling unref() again will have no effect.