An example of process.send vs worker.send.

cluster.send.js

var cluster = require('cluster')

if (cluster.isMaster) {
  console.log('Master started.')

  // If we're a master process, we can listen on the 'fork' event to get Worker instances.
  // We'll use it to communicate with that process.
  cluster.on('fork', function (worker) {
    // Now that we have a worker, we can listen on 'message' events that worker has sent.
    // These messages are sent via process.send, below, and are received only by the master.
    worker.on('message', function (msg) {
      console.log('Master received %j from worker %s.', msg, worker.id)
    })

    // If we want to send a message to a specific worker, we can use worker.send for that.
    worker.send('Welcome!')
  })

  // Now that's established, so let's fork a couple processes.
  cluster.fork()
  cluster.fork()

  // NOTE: If we call process.send from the master process, that won't go to ANY worker. It
  // will try to send the message to a parent process (of which there is none in this simple
  // example. If this paragraph doesn't make sense, please leave a comment and I'll provide
  // a separate example for this behaviour.
} else {
  console.log('Worker started.')

  // If we're a worker process, process.send can be used to communicate with the master process.
  setInterval(function () {
    process.send('Random message is random: ' + Math.random())
  }, 1000)

  // Additionally, we can listen on the same 'message' event (but on process instead of on a
  // Worker) to receive messages from the master.
  process.on('message', function (msg) {
    console.log('Worker received %j from the master.', msg)
  })
}

// So it's always .send() and .on('message'), but the master uses a Worker instance gotten from
// cluster.on('fork'), and the Worker uses the global process object.