Deploying to Heroku¶

This page explains the process we took to deploy Kvasir on Heroku. You can follow a similar approach for any third party hosting service but some of the particulars will be dependent on which service you chose to use.

Purpose¶

The point of deploying on Heroku was to validate that our system could easily be deployed. We wanted to make sure that all modules were included in our package.json file and that Kvasir could function outside the comfort of our development environment.

Our first attempt was less than glamorous. There were a lot of changes made in order to get the deployment to work, but we ended up with a better system overall. The idea for using environment variables instead of a configuration file came from this process.

Following the Tutorial¶

We are not the first people to deploy a NodeJS application to Heroku. Not even close.

There is a very good tutorial for deploying a NodeJS app on Heroku. This document will be more notes on the pitfalls that we encountered specific to Kvasir and what steps we took to climb out of them.

Procfile¶

The Heroku tutorial mentions that you can use a Procfile to specify how to launch the application; however, Heroku will also try and run commands present in the scripts section of the package.json file. Kvasir includes the commands for how to start the NodeJS server and build the front-end JavaScript library in the package.json file, so the Procfile is not necessary.

Setting Environment Variables¶

Heroku will only launch with files include in the git repository that you push to it. Assuming that you didn’t make any changes to Kvasir’s .gitignore file, then the .env file that includes all of the environment variable configuration, but is hidden from git, will not be included in the source code uploaded to Heroku. This presents a problem because Kvasir needs those configuration values, but the file that you put them in is no longer accessible.

The solution is to set environment variables on the Heroku server.

Depending on what version of Heroku toolbelt that you have installed on your machine you will use either heroku config:add or heroku config:set followed by the name of the environment variable and it’s corresponding value.

For example, to set the cookie secret:

>>> herkou config:set KVASIR_COOKIE_SECRET=YOUR_COOKIE_SECRET

You have to run this for each environment variable that you want to set, and the Heroku server will restart for each variable that is added or changed.

Scripting Makes Life Easier¶

If you already have a .env file defined (which you generally should during development), then you can use this Python script to set all of the environment variables on Heroku for you.

from subprocess import call

command = ['heroku', 'config:set', '']
with open(".env", "r") as f:
    for line in f:
        command[-1] = line.rstrip().replace(" ", '')
        call(command)

What this script does is open the .env file, and then for each line in that file, it will strip any trailing whitespace (including new lines) and remove any spaces from the line and send the command via the command line to heroku to add the environment variable and value.

The reason we have to remove any spaces is because spaces will mess up how the command line parses the command. If your KVASIR_MIDDLEWARE_URI has spaces in it, make sure you replace those with %20.

Testing Locally¶

Before deploying to Heroku, you should verify that everything works locally.

To do this, you go ahead and run:

>>> heroku local

This will emulate how the application will run on Heroku. If you are able to connect and perform actions on the local version, chances are very good that it will work when deployed to Heroku.