Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Zero-downtime reloads and requests load balancer based on distribute.
branch: master

This branch is 13 commits ahead, 63 commits behind LearnBoost:master

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.
example
lib
test
.gitignore
.travis.yml
README.md
package.json

README.md

This is a fork of learnboost's Up. Some addtional features and to match special needs at Resumable :)

Up Build Status

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

Features

  • Works with Node 0.6+
  • Works at the HTTP request level. It never drops requests or destroys Keep-Alive sockets while reloading.
  • Compatible with any HTTP server.
  • Built on distribute.

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');
});

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:

  • server (http.Server) server to accept connections on
  • module (String) absolute path to the module.
  • options (Object)
    • numWorkers: (Number): ms
    • workerTimeout: (Number): ms

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>

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.

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.

Something went wrong with that request. Please try again.