Cluster solution based on node's cluster module.

Cluster solution based on node's cluster module.


  • Restarts workers that die
  • Zero downtime reloading of application when receiving SIGHUP
  • Gracefully dies when receiving SIGQUIT
  • Correctly reloads code when deployed with capistrano


  • Overrides net._createServerHandle to track sockets open by master so they can be closed during graceful shutdown. If they are not closed then new clients will hang instead of receiving connection refused.
  • Overrides net.Server.prototype.listen in worker to ensure an error is generated if a port cannot be bound to


  • Not enough testing
  • Nodejs's cluster module doesn't appear to be fair (possibly why people have created their own nodejs http proxy balancers)




Example app.js

var ncluster = require('ncluster');
ncluster('./server.js', {workers: 5});

Example server.js

var http = require('http');
var http_server = http.createServer(function(req, res) {
  res.writeHead(200, {'Content-Type': 'text/plain'});
  res.end('Hello From Worker\n');


module.exports = http_server;