Testing a Phoenix application with GitLab CI/CD

Phoenix is a web development framework written in Elixir, which is a functional language designed for productivity and maintainability that runs on the Erlang VM. Erlang VM is really really fast and can handle very large numbers of simultaneous users.

That's why we're hearing so much about Phoenix today.

In this tutorial, we'll teach you how to set up GitLab CI/CD to build and test a Phoenix application.

We assume that you know how to create a Phoenix app, run tests locally, and how to work with Git and GitLab UI.

Introduction

What is Phoenix?

Phoenix is a web development framework written in Elixir very useful to build fast, reliable, and high-performance applications, as it uses Erlang VM.

Many components and concepts are similar to Ruby on Rails or Python's Django. High developer productivity and high application performance are only a few advantages on learning how to use it. Working on the MVC pattern, it's was designed to be modular and flexible. Easy to mantain a growing app is a plus.

Phoenix can run in any OS where Erlang is supported:

Ubuntu
CentOS
Mac OS X
Debian
Windows
Fedora
Raspbian

Check the Phoenix learning guide for more information.

What is Elixir?

Elixir is a dynamic, functional language created to use all the maturity of Erlang (30 years old!) in these days, in an easy way. It has similarities with Ruby, specially on sintax, so Ruby developers are quite excited with the rapid growing of Elixir. A full-stack Ruby developer can learn how to use Elixir and Phoenix in just a few weeks!

In Elixir we have a command called mix, which is a helper to create projects, testing, run migrations and much more. We'll use it later on in this tutorial.

Check the Elixir documentation for more information.

Requirements

To follow this tutorial, you'll need to have installed:

Elixir installation instructions
Phoenix Framework installation instructions
PostgreSQL (if you need to use MySQL server, check Phoenix instructions)

Create a new Phoenix project

Open your terminal and go to the directory you wish to create your project. You don't need to create an empty directory for the project's files, because the mix command will do it for us.

When we call mix command, we'll pass two arguments:

The task we want it to run: phoenix.new
And the parameter phoenix.new requires, which is the name of the new project. In this case, we're calling it hello_gitlab_ci, but you're free to set your own name:

mix phoenix.new hello_gitlab_ci

When asked, answer Y to fetch and install dependencies.

If everything went fine, you'll get an output like this:

Now, our project is located inside the directory with the same name we pass to mix command, for example, ~/GitLab/hello_gitlab_ci. If we take a look at the directory, we'll see the Phoenix files and the dependencies needed to run.

Initialize the PostgreSQL database

By default, Phoenix requires a PostgreSQL database to store whatever we need to store in our app. In this case, we'll only create an empty database.

First, we need to navigate to our recently created project's directory, and then execute again mix. This time, mix will receive the parameter ecto.create, which is the task to create our new database. Ecto is the database wrapper for Elixir.

When we do run mix the first time after creating our project, it will compile our files to bytecode, which will be interpreted by Erlang VM. In the next times, it will only compile our changes.

Run the commands below to create our empty database:

cd hello_gitlab_ci
mix ecto.create

We expect to see this output at the end of the command:

Generated hello_gitlab_ci app
The database for HelloGitlabCi.Repo has been created

Note: Phoenix assumes that our PostgreSQL database will have a postgres user account with the correct permissions and a password of postgres. If it's not your case, check Ecto's instructions.

Start Phoenix server

Now, it's time to see if everything we did until now went well. We'll call mix again, this time with phoenix.server parameter, which will start Phoenix's HTTP Server.

mix phoenix.server

This will be the output to this command:

[info] Running HelloGitlabCi.Endpoint with Cowboy using http://localhost:4000
23 May 11:44:35 - info: compiling
23 May 11:44:37 - info: compiled 6 files into 2 files, copied 3 in 9.8 sec

Now, we have our app running locally. We can preview it directly on our browser. Let's open localhost:4000 to see our Phoenix Framework welcome page. If the link do not work, open 127.0.0.1:4000 instead and later, configure your OS to point localhost to 127.0.0.1.

Great, now we have a local Phoenix Server running our app.

Locally, our application is running in an iex session. iex stands for Interactive Elixir. In this interactive mode, we can type any Elixir expression and get its result. To exit iex, we need to press Ctrl+C twice. So, when we need to stop the Phoenix server, we have to hit Ctrl+C twice.

Introducing GitLab CI/CD

With GitLab, we can manage our development workflow, improve our productivity, track issues, perform code review, and much more from a single platform. With GitLab CI/CD, we can be much more productive, because every time we, or our co-workers push any code, GitLab CI/CD will build and test the changes, telling us in realtime if anything goes wrong.

Certainly, when our application starts to grow, we'll need more developers working on the same project and this process of building and testing can easely become a mess without proper management. That's also why GitLab CI/CD is so important to our application. Every time someone pushes its code to GitLab, we'll quickly know if their changes broke something or not. We don't need to stop everything we're doing to test manually and locally every change our team does.

Let's see this in practice.

Adjusting Phoenix configuration

Now, we need to adjust our Phoenix configuration before configuring GitLab CI/CD. There is a directory (config) in your Phoenix project that contains a configuration file for every environment it can run. Since we will work with a single environment, we'll edit just the test configuration file (test.exs).

But, why do we need to adjust our configuration? Well, GitLab CI/CD builds and tests our code in one isolated virtual machine, called Runner, using Docker technology. In this Runner, GitLab CI/CD has access to everything our Phoenix application need to run, exactly as we have in our localhost, but we have to tell GitLab CI/CD where to create and find this database using system variables. This way, GitLab CI/CD will create our test database inside the Runner, just like we do when running our Phoenix in our localhost.

Open hello_gitlab_ci/config/test.exs on your favorite code editor

Go to Configure your database session and edit the block to include System.get_env:

# Configure your database
config :hello_gitlab_ci, HelloGitlabCi.Repo,
  adapter: Ecto.Adapters.Postgres,
  username: System.get_env("POSTGRES_USER") || "postgres",
  password: System.get_env("POSTGRES_PASSWORD") || "postgres",
  database: System.get_env("POSTGRES_DB") || "hello_gitlab_ci_test",
  hostname: System.get_env("POSTGRES_HOST") || "localhost",
  pool: Ecto.Adapters.SQL.Sandbox

We'll need these system variables later on.

Create an empty file named .gitkeep into hello_gitlab_ci/priv/repo/migrations

As our project is still fresh, we don't have any data on our database, so, the migrations directory will be empty. Without .gitkeep, git will not upload this empty directory and we'll got an error when running our test on GitLab.

Note: If we add a folder via the GitLab UI, GitLab itself will add the .gitkeep to that new dir.

Now, let's run a local test and see if everything we did didn't break anything.

Testing

Earlier, when we created our project, we ran mix phoenix.new. This task created everything a Phoenix application needed, including some unit tests into hello_gitlab_ci/test directory.

Let's run a new task with mix to run those tests for us. This time, the parameter expected is test. We can add --trace parameter for debugging purposes.

In your terminal, navigate to the directory hello_gitlab_ci and run:

mix test

Our expected result is this:

....

Finished in 0.7 seconds
4 tests, 0 failures

Randomized with seed 610000

Our test was successfull. It's time to push our files to GitLab.

Configuring CI/CD Pipeline

The first step is to create a new file called .gitlab-ci.yml in hello_gitlab_ci directory of our project.

The fastest and easiest way to do this, is to click on Set up CI on project's main page:
On next screen, we can select a template ready to go. Click on Apply a GitLab CI/CD Yaml template and select Elixir:

This template file tells GitLab CI/CD about what we wish to do every time a new commit is made. However, we have to adapt it to run a Phoenix app.
The first line tells GitLab what Docker image will be used.

Remember when we learn about Runners, the isolated virtual machine where GitLab CI/CD build and test our application? This virtual machine must have all dependencies to run our application. This is where a Docker image is needed. The correct image will provide the entire system for us.

As a suggestion, you can use trenpixster's elixir image, which already has all dependencies for Phoenix installed, such as Elixir, Erlang, NodeJS and PostgreSQL:
```
image: trenpixster/elixir:latest
```
At services session, we'll only use postgres, so we'll delete mysql and redis lines:
```
services:
  - postgres:latest
```
Now, we'll create a new entry called variables, before before_script session:
```
variables:
  POSTGRES_DB: hello_gitlab_ci_test
  POSTGRES_HOST: postgres
  POSTGRES_USER: postgres
  POSTGRES_PASSWORD: "postgres"
  MIX_ENV: "test"
```
Here, we are setting up the values for GitLab CI/CD authenticate into PostgreSQL, as we did on config/test.exs earlier.
In before_script session, we'll add some commands to prepare everything to the test:
```
before_script:
  - apt-get update && apt-get -y install postgresql-client
  - mix local.hex --force
  - mix deps.get --only test
  - mix ecto.create
  - mix ecto.migrate
```
It's important to install postgresql-client to let GitLab CI/CD access PostgreSQL and create our database with the login information provided earlier. More important is to respect the identation, to avoid syntax errors when running the build.
And finally, we'll let mix session intact.

Let's take a look at the completed file after the editions:

image: trenpixster/elixir:latest

services:
  - postgres:latest

variables:
  POSTGRES_DB: test_test
  POSTGRES_HOST: postgres
  POSTGRES_USER: postgres
  POSTGRES_PASSWORD: "postgres"
  MIX_ENV: "test"

before_script:
  - apt-get update && apt-get -y install postgresql-client
  - mix deps.get
  - mix ecto.create
  - mix ecto.migrate

mix:
  script:
  - mix test

For safety, we can check if we get any syntax errors before submiting this file to GitLab. Copy the contents of .gitlab-ci.yml and paste it on GitLab CI/CD Lint tool. Please note that this link will only work for logged in users.

Watching the build

I don't know about you, but I love to watch that black screen being filled with compilation output. With this, I can feel the happiness of something I made working correctly. On localhost it's easy to watch our build, but on GitLab, is it possible? Yes!

Let's go to Pipelines and see GitLab doing the job. Just click on Pipelines to find the actual running build job.

Click on build's ID to watch the entire process. If everything went as expected, we can wait for the Build succeeded at the end of the process! :)

$ mix test
....

Finished in 0.3 seconds
4 tests, 0 failures

Randomized with seed 206909
Build succeeded

If we take a look at the project's main page on the GitLab UI, we can see the status of the last build made by GitLab CI/CD.

Time to show the world our green build badge! Navigate to your project's Settings > CI/CD and expand General pipelines settings. Scroll down to Pipeline status and copy the markdown code for your badge. Paste it on the top of your README.md file, to let people outside of our project see if our latest code is running without errors.

When we finish this edition, GitLab will start another build and show a build running badge. It is expected, after all we just configured GitLab CI/CD to do this for every push! But you may think "Why run build and tests for simple things like editing README.md?" and it is a good question. For changes that don't affect your application, you can add the keyword [ci skip] to commit message and the build related to that commit will be skipped.

In the end, we finally got our pretty green build succeeded badge! By outputting the result on the README file, it shows to whoever lands on your project's page that your code is up-to-date and working properly.

Conclusion

When we have a growing application with many developers working on it, or when we have an open source project being watched and contributed by the community, it is really important to have our code permanently working. GitLab CI/CD is a time saving powerfull tool to help us mantain our code organized and working.

As we could see in this post, GitLab CI/CD is really really easy to configure and use. We have many other reasons to keep using GitLab CI/CD. The benefits to our teams will be huge!