Skip to content
Production-ready, scaling-ready Node + Express on Heroku with a message broker (amqp) and state (mongo).
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


Running Node all the way from development to production on Heroku.

Check it out at


Local dependencies


$ brew install mongodb rabbitmq
$ brew services start mongodb
$ brew services start rabbitmq
$ npm install


  1. npm start
  2. http://localhost:5000


Deploying is easy - just use the Heroku Button:


If you'd rather clone locally and then deploy through the CLI, you can do that too:

git clone
cd node-articles-nlp

heroku create

heroku addons:add mongohq
heroku addons:add cloudamqp

heroku config:set NODE_ENV=production
heroku config:set VIEW_CACHE=true
heroku config:set THRIFTY=true

git push heroku master
heroku open


Environment variables are mapped to a config object in lib/config.js. This provides reasonable defaults as well as a layer of generalization (process.env.MONGOHQ_URL => config.mongo_url).

You can locally override the defaults by adding variables to a .env file.


The app is separated into two tiers:

This enables horizontally scaling both web traffic and long-running jobs.

On Heroku

The default deploy configuration includes THRIFTY=true, which starts the app in single-dyno mode to avoid charges. With THRIFTY=true, the web process handles both http requests and queued jobs. Keep in mind that this is a specific setting for this app, as THRIFTY is not a standard Heroku configuration.

Of course, a production app should never run in a single instance or make users wait for worker processes. When you're ready to test in staging or deploy to production, you can scale beyond single-dyno mode:

heroku config:unset THRIFTY
heroku ps:scale web=2 worker=2


npm start runs node-foreman, which will check the Procfile and start a single web process and a single worker process.

To test that your app behaves correctly when clustered in multiple processes, you can specify process scales to node-forman and set CONCURRENCY=4 in a local .env file.


Writing maintainable Node apps is all about separating concerns into small, well-defined modules. This barebones app has three distinct components with their own responsibilities:


The business logic is all in lib/app. This module orchestrates and provides a facade for the underlying MongoDB database and the RabbitMQ job queue.


The user-facing portion of the project lies in lib/web. This module is responsible for providing an http interface and routing requests. It shows things and relies on an App instance to do things.


The background processes run through lib/worker. This module is tiny - it just instantiates an App instance to process the job queue.

You can’t perform that action at this time.