LearnBoost / up

Zero-downtime reloads and requests load balancer based on distribute.
540 stars 73 forks source link

NOTE: up has been renamed in npm as "up-time" https://npmjs.org/package/up-time

Up

Zero-downtime reloads built on top of the distribute load balancer.

Simply running

$ up --port 80 --watch my-http-server.js

Will start my-http-server.js on port 80, then reload it with no downtime when files change in the working directory.

This project is deprecated!

Features

Setup

Make sure you structure your code so that your http server lives in a separate module that can be required.

server.js

module.exports = http.Server(function (req, res) {
  res.writeHead(200);
  res.end('Hello World');
});

A) CLI

To get the up command, make sure to install with the -g flag:

$ npm install -g up

Usage: up [options]

The up command accepts the following options:

B) JavaScript API

var up = require('up')
  , master = http.Server().listen(3000)

// initialize up
var srv = up(master, __dirname + '/server');

process.on('SIGUSR2', function () {
  srv.reload();
});

require('up') exports the UpServer constructor, which takes three parameters:

Middleware

An UpServer inherits from a Distributor, which means you can use() any distribute middleware.

The main difference is that the "default handler" of up (ie: the last function in the middleware chain) is the one that executes the round-robin load balancing.

Reloading

To reload the workers, call srv.reload(). In the example above and CLI, this is called by sending the SIGUSR2 signal:

$ kill -s SIGUSR2 <process id>

If you're running with up CLI, this command is output to stderr for your convenience.

The CLI tool also auto-reloads if you pass the --watch option and a file changes in the working directory.

Strategy

  1. An up server starts with an arbitrary number of workers, which defaults to the number of CPUs.
  2. When a reload instruction is received, it spawns an identical number of workers.
  3. Upon the first of those workers binding to a port, any subsequent requests are sent to that worker, and all the workers containing old code are discarded.
  4. The discarded workers could have been processing requests, so they only truly die after the configured workerTimeout, which defaults to 10 minutes in production. This means that if a user was uploading a file, his request will be processed without interruptions.
  5. As other workers bind and become available, they join the round-robin round.

Worker readiness

By default up assume that new workers are ready for new connections, immediately after they have been required. This can be changed by setting assumeReady to false, on the options object when initializing the up server through the JavaScript API.

The worker then needs to tell up, when it's ready, like this:

var up = require('up');
// Dummy async event
setTimeout(function(){
    up.ready();
}, 1000);

Credits

(The MIT License)

Copyright (c) 2011 Guillermo Rauch <guillermo@learnboost.com>

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the 'Software'), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.