How to Dockerize a Rails application (“hello world” version)

by Jason Swett,

Reasons to Dockerize a Rails application

I had to hear Docker explained many times before I finally grasped why it’s useful.

I’ll explain in my own words why I think Dockerizing my applications is something worth exploring. There are two benefits that I recognize: one for the development environment and one for the production environment.

Development environment benefits

When a new developer joins a team, that developer has to get set up with the team’s codebase(s) and get a development environment running. This can often be time-consuming and tedious. I’ve had experiences where getting a dev environment set up takes multiple days. If an application is Dockerized, spinning up a dev environment can be as simple as running a single command.

In addition to simplifying the setup of a development environment, Docker can simplify the running of a development environment. For example, in addition to running the Rails server, my application also needs to run Redis and Sidekiq. These services are listed in my Procfile.dev file, but you have to know that file is there, and you need to know to start the app using foreman start -f Procfile.dev. With Docker you can tell it what services you need to run and Docker will just run the services for you.

Production environment benefits

There are a lot of different ways to deploy an application to production. None of them is particularly simple. As of this writing, my main production application is deployed to AWS using Ansible for infrastructure management. This is nice in many ways but it’s also somewhat duplicative. I’m using one (currently manual) way to set up my development environment and then another (codified, using Ansible) ways to set up my production environment.

Docker allows me to set up both my development environment and production environment the same way, or at least close to the same way. Once I have my application Dockerized I can use a tool like Kubernetes to deploy my application to any cloud provider without having to do a large amount of unique infrastructure configuration myself the way I currently am with Ansible. (At least that’s my understanding. I’m not at the stage of actually running my application in production with Docker yet.)

What we’ll be doing in this tutorial

In this tutorial we’ll be Dockerizing a Rails application using Docker and a tool called Docker Compose.

The Dockerization we’ll be doing will be the kind that will give us a development environment. Dockerizing a Rails app for use in production hosting will be a separate later tutorial.

My aim for this tutorial is to cover the simplest possible example of Dockerizing a Rails application. What you’ll get as a result is unfortunately not robust enough to be usable as a development environment as-is, but will hopefully serve as a good exercise to build your Docker confidence and to serve as a good jumping-off point for creating a more robust Docker configuration.

In this example our Rails application will have a PostgreSQL database and no other external dependencies. No Redis, no Sidekiq. Just a database.

Prerequisites

I’m assuming that before you begin this tutorial you have both Docker and Docker Compose installed. I’m assuming you’re using a Mac. Nothing else is required.

If you’ve never Dockerized anything before, I’d recommend that you check out my other post, How to Dockerize a Sinatra application, before digging into this one. The other post is simpler because there’s less stuff involved.

Fundamental Docker concepts

Let’s say I have a tiny Dockerized Ruby (not Rails, just Ruby) application. How did the application get Dockerized, and what does it mean for it to be Dockerized?

I’ll answer this question by walking sequentially through the concepts we’d make use of during the process of Dockerizing the application.

Images

When I run a Dockerized application, I’m running it from an image. An image is kind of like a blueprint for an application. The image doesn’t actually do anything, it’s just a definition.

If I wanted to Dockerize a Ruby application, I might create an image that says “I want to use Ruby 2.7.1, I have such-and-such application files, and I use such-and-such command to start my application”. I specify all these things in a particular way inside a Dockerfile, which I then use to build my Docker image.

Using the image, I’d be able to run my Dockerized application. The application would run inside a container.

Containers

Images are persistent. Once I create an image, it’s there on my computer (findable using the docker images command) until I delete it.

Containers are more ephemeral. When I use the docker run command to run an image, part of what happens is that I get a container. (Containers can be listed using docker container ls.) The container will exist for a while and then, when I kill my docker run process, the container will go away.

The difference between Docker and Docker Compose

One of the things that confused me in other Rails + Docker tutorials was the usage of Docker Compose. What is Docker Compose? Why do we need to use it in order to Dockerize Rails?

Docker Compose is a tool that lets you Dockerize an application that’s composed of multiple containers.

When in this example we Dockerize a Rails application that uses PostgreSQL, we can’t use just one image/container for that. We have to have one container for Rails and one container for PostgreSQL. Docker Compose lets us say “hey, my application has a Rails container AND a PostgreSQL container” and it lets us say how our various containers need to talk to each other.

The files involved in Dockerizing our application

Our Dockerized Rails application will have two containers: one for Rails and one for PostgreSQL. The PostgreSQL container can mostly be grabbed off the shelf using a base image. Since certain container needs are really common—e.g. a container for Python, a container for MySQL, etc.—Docker provides images for these things that we can grab and use in our application.

For our PostgreSQL need, we’ll grab the PostgreSQL 11.5 image from Docker Hub. Not much more than that is necessary for our PostgreSQL container.

Our Rails container is a little more involved. For that one we’ll use a Ruby 2.7.1 image plus our own Dockerfile that describes the Rails application’s dependencies.

All in all, Dockerizing our Rails application will involve two major files and one minor one. An explanation of each follows.

Dockerfile

The first file we’ll need is a Dockerfile which describes the configuration for our Rails application. The Dockerfile will basically say “use this version of Ruby, put the code in this particular place, install the gems using Bundler, install the JavaScript dependencies using Yarn, and run the application using this command”.

You’ll see the contents of the Dockerfile later in the tutorial.

docker-compose.yml

The docker-compose.yml file describes what our containers are and how they’re interrelated. Again, we’ll see the contents of this file shortly.

init.sql

This file plays a more minor role. In order for the PostgreSQL part of our application to function, we need a user with which to connect to the PostgreSQL instance. The only way to have a user is for us to create one. Docker allows us to have a file called init.sql which will execute once per container, ever. That is, the init.sql will run the first time we run our container and never again after that.

Dockerizing the application

Start from this repo called boats.

$ git clone git@github.com:jasonswett/boats.git

The master branch is un-Dockerized. You can start here and Dockerize the app yourself or you can switch to the docker branch which I’ve already Dockerized.

Dockerfile

Paste the following into a file called Dockerfile and put it right at the project root.

# Use the Ruby 2.7.1 image from Docker Hub
# as the base image (https://hub.docker.com/_/ruby)
FROM ruby:2.7.1

# Use a directory called /code in which to store
# this application's files. (The directory name
# is arbitrary and could have been anything.)
WORKDIR /code

# Copy all the application's files into the /code
# directory.
COPY . /code

# Run bundle install to install the Ruby dependencies.
RUN bundle install

# Install Yarn.
RUN curl -sS https://dl.yarnpkg.com/debian/pubkey.gpg | apt-key add -
RUN echo "deb https://dl.yarnpkg.com/debian/ stable main" | tee /etc/apt/sources.list.d/yarn.list
RUN apt-get update && apt-get install -y yarn

# Run yarn install to install JavaScript dependencies.
RUN yarn install --check-files

# Set "rails server -b 0.0.0.0" as the command to
# run when this container starts.
CMD ["rails", "server", "-b", "0.0.0.0"]

docker-compose.yml

Create another file called docker-compose.yml. Put this one at the project root as well.

# Use the file format compatible with Docker Compose 3.8
version: "3.8"

# Each thing that Docker Compose runs is referred to as
# a "service". In our case, our Rails application is one
# service ("web") and our PostgreSQL database instance
# is another service ("database").
services:

  database:
    # Use the postgres 11.5 base image for this container.
    image: postgres:11.5

    volumes:
      # We need to tell Docker where on the PostgreSQL
      # container we want to keep the PostgreSQL data.
      # In this case we're telling it to use a directory
      # called /var/lib/postgresql/data, although it
      # conceivably could have been something else.
      #
      # We're associating this directory with something
      # called a volume. (You can see all your Docker
      # volumes by running +docker volume ls+.) The name
      # of our volume is db_data.
      - db_data:/var/lib/postgresql/data

      # This copies our init.sql into our container, to
      # a special file called
      # /docker-entrypoint-initdb.d/init.sql. Anything
      # at this location will get executed one per
      # container, i.e. it will get executed the first
      # time the container is created but not again.
      #
      # The init.sql file is a one-line that creates a
      # user called (arbitrarily) boats_development.
      - ./init.sql:/docker-entrypoint-initdb.d/init.sql

  web:
    # The root directory from which we're building.
    build: .

    # This makes it so any code changes inside the project
    # directory get synced with Docker. Without this line,
    # we'd have to restart the container every time we
    # changed a file.
    volumes:
      - .:/code:cached

    # The command to be run when we run "docker-compose up".
    command: bash -c "rm -f tmp/pids/server.pid && bundle exec rails s -p 3000 -b '0.0.0.0'"

    # Expose port 3000.
    ports:
      - "3000:3000"

    # Specify that this container depends on the other
    # container which we've called "database".
    depends_on:
      - database

    # Specify the values of the environment variables
    # used in this container.
    environment:
      RAILS_ENV: development
      DATABASE_NAME: boats_development
      DATABASE_USER: boats_development
      DATABASE_PASSWORD: 
      DATABASE_HOST: database

# Declare the volumes that our application uses.
volumes:
  db_data:

init.sql

This one-liner is our third and final Docker-related file to add. It will create a PostgreSQL user for us called boats_development. Like the other two files, this one can also go at the project root.

CREATE USER boats_development SUPERUSER;

config/database.yml

We’re done adding our Docker files but we still need to make one change to the Rails application itself. We need to modify the Rails app’s database configuration so that it knows it needs to be pointed at a PostgreSQL instance running in the container called database, not the same container the Rails app is running in.

default: &default
  adapter: postgresql
  encoding: unicode
  database: <%= ENV['DATABASE_NAME'] %>
  username: <%= ENV['DATABASE_USER'] %>
  password: <%= ENV['DATABASE_PASSWORD'] %>
  port: 5432
  host: <%= ENV['DATABASE_HOST'] %>
  pool: <%= ENV.fetch("RAILS_MAX_THREADS") { 5 } %>
  timeout: 5000

development:
  <<: *default

test:
  <<: *default

production:
  <<: *default

Building and running our Dockerized application

Run the following command to build the application we’ve just described in our configuration files.

$ docker-compose build

Once that has successfully completed, run docker-compose up to run our application’s containers.

$ docker-compose up

The very last step before we can see our Rails application in action is to create the database, just like we would if we were running a Rails app without Docker.

$ docker-compose run web rails db:create
$ docker-compose run web rails db:migrate

The docker-compose run command is what we use to run commands inside a container. Running docker-compose run web means “run this command in the container called web“.

Finally, open http://localhost:3000 to see your Dockerized Rails application running in the browser.

$ open http://localhost:3000

Congratulations. You now have a Dockerized Rails application.

If you followed this tutorial and it didn’t work for you, please leave a comment with your problem, and if I can, I’ll help troubleshoot. Good luck.

26 thoughts on “How to Dockerize a Rails application (“hello world” version)

  1. Andrew

    Very clear explanation of Docker terms. But not sure if I’m missing something, as I find Docker confusing and poorly documented, but isn’t this unsuitable for development as the code dir is never mounted? The tiniest change would result in a full image rebuild?

    Also I know you were going for the simplest example but I think of the huge amount of bandwidth that RubyGems has to deal with because so many Docker configs don’t cache the gems. I think you at least need to do the “COPY Gemfile” silly dance first to have a stable “layer”.

    Reply
  2. Clemens

    What you’ve described is essentially the easy part (IMO). That’s great, but the difficult thing (again: IMO) is then setting up things like Capybara with WebDriver in a sensible way (i.e. Chrome in a separate container etc.). Maybe that would be a good follow-up tutorial?

    Cheers!

    Reply
    1. Jason Swett Post author

      Thanks! I agree that it’s maybe the relatively easy part, although for me, getting all this stuff figured out was actually pretty hard, for whatever it’s worth.

      I agree that getting Capybara etc. set up would be a good follow-up. I actually recently did that for a production application.

      Reply
  3. Yves Siegrist

    Gotta love the autocompletion of “Shit” and “Fucker” at 6:06.

    Thanks for the quick explanation, I do agree with Andrew on the layer which should be created so you don’t always have to reinstall all the gems when changing a model for example.

    Reply
  4. Samuel Rowe

    Great article Jason 🙂

    A couple of cool things you might want to checkout as next steps ( full disclosure I work for Microsoft as a Solution Architect in the UK )

    1) have you seen VSCode Dev Containers, where you develop inside a container. The dev container docker file is kept with the repo so getting a new dev environment setup is really fast and repeatable. Sharing parts of the docker file between the dev container and the production container is easy too! – https://code.visualstudio.com/docs/remote/containers

    2) Hosting wise, if you had an app with one or two containers wouldn’t you be better to host on a PasS platform such as AWS Elastic Beanstalk or Azure App Service? – https://azure.microsoft.com/en-us/services/app-service/containers/ and see this link for more details https://docs.microsoft.com/en-us/azure/app-service/tutorial-multi-container-app

    Let me know if I can help any, easy to find me on Twitter or LinkedIn

    Reply
  5. Christian

    FWIW

    I’ve been able to avoid Docker thus far to run my many local Rails apps by leveraging https://asdf-vm.com

    It’ll handle each apps unique Ruby, Node, Yarn, Redis, and Postgres version dependencies in a unified way.

    It is worth noting that I use the MacOS Postgres.app to handle running different Postgres versions and all apps use DelayedJob currently so I don’t have a Redis dependency.

    I just started setting up my Linux laptop, which I use when I’m on the go and as a backup, and ASDF is working well there so far too. In that environment I am using the ASDF Postgres plugin.

    Reply
  6. Kamil Krzywicki

    I found precompiling and serving assets in a “proper way” quite challenging. Production setups should precompile assets and then somehow serve them using web server, not puma or other rails application server.
    Another challenge was how to configure many environments (test, stage, prod, dev). My CI builds one image for rails environment docker, and everything is configured by injecting configuration from container orchestration tool (docker compose, kubernetes).

    Reply
  7. Maicol

    We used docker for the same reasons and had to rollback it because its slowness on Mac, now we have one command that setups everything in your machine, I know it isn’t the best but I prefer waste time there that each time running a test.

    Reply
  8. KD

    Do you have thoughts on Continuous Deployment with Rails/Docker? I’ve found a lot of content on Rails/Docker seems to avoid this, and it’s quite hard to find info on best practice. E.g. after migrating from Heroku into a Dockerized world, how do you replicate the deploy-rollback scenarios.

    I’ve worked something up that looks at log events from within the containers to know when a deploy has succeeded (to then kill old containers), but it still feels flaky to me.

    Reply
  9. Carlos Herrera

    I have used Docker/Rails to make MVPs and scale them as per requirements. I have actually built great, simple yet powerful solutions.
    Really great article Jason! Keep sharing such amazing ones.

    Reply
  10. Jetson Davis

    Hey Jason,

    Just tried to follow you recipe, got this, please advise??

    Creating boats_database_1 … done
    Creating boats_web_run … done
    error Couldn’t find an integrity file
    error Found 1 errors.

    ========================================
    Your Yarn packages are out of date!
    Please run `yarn install –check-files` to update.
    ========================================

    To disable this check, please change `check_yarn_integrity`
    to `false` in your webpacker config file (config/webpacker.yml).

    yarn check v1.22.5
    info Visit https://yarnpkg.com/en/docs/cli/check for documentation about this command.

    Reply
    1. Nick Mixdorf

      Before you run `docker-compose run web rails db:create`

      Run this to check you yarn install

      docker-compose run web yarn install –check-files

      Reply
  11. Arthur

    Is this recipe for development or deployment though? How does it deal with hot code reloading? Isn’t it better to have a volume which will be attached to container to not copy the code all the time? The same is true about the volume for gems

    Reply
    1. Jason Swett Post author

      Development, and it’s not even really intended to be used for development, it’s just an example as an illustration.

      I don’t think I understand your question about copying code. The files are shared via a synced filesystem between the container and host.

      Reply
  12. Facundo

    Hey Jason, great tutorial!

    Question, besed on your Dockerfile code, you copy the entire project folder and then you run bundle install, but with this configuration you’ll need to install all the gems everytime the code change because the `COPY . /code` layer will be invalid, isn’t it? I think you should first `COPY Gemfile* /code` then `RUN bundle install` and then copy the entire folder, make sense?

    Reply
  13. Will

    Alright Jason, great post!

    Spinning up multiple containers for a rails project is a daunting task when you are new to Docker!

    I was smacking my head against the wall for a while but this post helped me a lot.

    Thank you!

    Will

    Reply
  14. docljn

    Thank you for this really useful tutorial. I was bitten by the mimemagic issue when I tried to follow the dockerizing instructions, but managed to find a way around it by using a newer version of Rails.

    Reply
  15. Megan

    Hi Jason, I found your article to be really helpful. If I just try to dockerize an existed Ruby app, can I leave out the Postgre database part? I have to dockerize an existed Ruby app as my homework assigment.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *