Continuous Integration for Rails project using Github Actions

Intro

With the release of Github Actions, we experimented with it to replace our current Continuous Integration (CI) process. This post describes the steps we took in order to do so.

Our CI includes of 3 checks:

1. specs
2. rubocop check
3. linter check for swagger docs

These checks run every time a developer pushes a commit or creates a Pull Requests (PR). The aim of this post is to give detailed description of the process to perform those jobs with Github Actions.

Defining a Workflow

Let’s start with creating a new workflow in Github Actions that will perform those tasks. In your root rails project:

mkdir -p .github/workflows
touch .github/workflows/main.yml

The main.yml file is where we define our workflow for CI. We name our workflow CI, and then list the name of events which will trigger our workflow.

name: CI
on: [push, pull_request]

We then describe the jobs that we want to run in the workflow. We can describe more than one job and each of them runs in parallel. Here we describe 3 different jobs for the aforementioned 3 checks.

name: CI
on: [push, pull_request]
jobs:
  specs:
  rubocop:
  swagger:

Defining a Job

Every job needs an instance of virtual host machine to run on. This is specified by statement runs-on. We can then define the sequential tasks that we want to perform inside this machine with steps statement.

name: CI
on: [push, pull_request]
jobs:
  specs:
    name: specs
    runs-on: ubuntu-latest
    steps:
    - name Install Libraries
      run: |
        sudo apt-get update
        sudo apt-get install -y postgresql-client libpq-dev

In the above workflow we defined a job named specs, that runs on ubuntu-latest instance. Inside the instance it installs package postgresql-client and libpq-dev. This is required for configuring the postgres database required for specs.

Services

Services can be used to create additional containers for a job or steps. In our case, we are using it to spin off postgres service.

name: CI
on: [push, pull_request]
jobs:
  specs:
    name: specs
    runs-on: ubuntu-latest
    services:
      postgres:
        image: postgres:10.8
        ports:
        - 5432:5432
    steps:
    - name Install Libraries
      run: |
        sudo apt-get update
        sudo apt-get install -y postgresql-client libpq-dev

Once we have our postgres service up and running, and postgres client installed, we will create test database.

name: CI
on: [push, pull_request]
jobs:
  specs:
    name: specs
    runs-on: ubuntu-latest
    services:
      postgres:
        image: postgres:10.8
        ports:
        - 5432:5432
    steps:
    - name Install Libraries
      run: |
        sudo apt-get update
        sudo apt-get install -y postgresql-client libpq-dev
    - name: Configure databases
      run: |
        echo "Postgres"
        psql -h localhost -c 'create database "test-database";' -U postgres

Actions by Github

Next thing we will do is checkout our rails app code. Github provides many official, ready to use actions, under its verified account, and one of those is actions/checkout. We can use this actions directly with the statement uses. Similarly, another actions actions/setup-ruby is used to checkout the ruby version that we want to use for our rails app.

name: CI
on: [push, pull_request]
jobs:
  specs:
    name: specs
    runs-on: ubuntu-latest
    services:
      postgres:
        image: postgres:10.8
        ports:
        - 5432:5432
    steps:
    - name Install Libraries
      run: |
        sudo apt-get update
        sudo apt-get install -y postgresql-client libpq-dev
    - name: Configure databases
      run: |
        echo "Postgres"
        psql -h localhost -c 'create database "test-database";' -U postgres
    - name: Checkout code
      uses: actions/checkout@v1

    - name: Set up Ruby
      uses: actions/setup-ruby@v1
      with:
        ruby-version: 2.6.3

Once the code has been checked out and correct ruby version is setup, we install the gems using bundler and then run the specs.

name: CI
on: [push, pull_request]
jobs:
  specs:
    name: specs
    runs-on: ubuntu-latest
    services:
      postgres:
        image: postgres:10.8
        ports:
        - 5432:5432
        env:
          POSTGRES_PASSWORD: ""
    steps:
    - name: Install libraries
      run: |
        sudo apt-get update
        sudo apt-get install -y postgresql-client libpq-dev
    - name: Configure databases
      run: |
        echo "Postgres"
        psql -h localhost -c 'create database "test-database";' -U postgres
    - uses: actions/checkout@v1

    - name: Set up Ruby
      uses: actions/setup-ruby@v1
      with:
        ruby-version: 2.6.5

    - name: Install bundler and gems
      run: |
        gem install bundler --no-document
        bundle config GITHUB__COM $GITHUB_ACCESS_TOKEN
        bundle install --jobs 4 --retry 3
      env:
        GITHUB_ACCESS_TOKEN: ${{ secrets.GITHUB_ACCESS_TOKEN }} 

    - name: Run tests
      run: |
        pg_config --version
        bin/rails db:schema:load RAILS_ENV=test
        bin/rspec

Secrets and Environment Variables

One interesting feature in the above yaml is the use of secrets. We can define the secrets in the settings section of our Github repository. More about it can be read here. The secrets defined in the Github repo, in our case, GITHUB_ACCESS_TOKEN, can be accessed inside our job instance. We used it to setup environment variable, which is passed into the step container to install gems.

env:
  GITHUB_ACCESS_TOKEN: ${{ secrets.GITHUB_ACCESS_TOKEN }} 

Environment variables can be set for each individual steps. Github also sets many environmental variables by default, which are accessible in every step. See here.

We follow the similar structure to run our other two jobs rubocop and swagger.

Slack Notification

There was one last function missing, slack notification. We wanted to be notified in the slack if our CI passed or failed. In order to notify when all the CI checks were success, we made use of workflow syntax needs. This statement defines the list of jobs that needs to be completed successfully before the described job will run. We used action rtCamp/actions-slack-notify, in order to notify the slack webhook.

slack_success:
    name: slack
    runs-on: ubuntu-latest
    needs: [swagger, rubocop, specs]
    steps:
      - uses: actions/checkout@v1
      - name: Success Notify
        uses: rtCamp/action-slack-notify@master
        env:
          SLACK_USERNAME: "Action Bot"
          SLACK_ICON: "https://github.com/freeletics.png"
          SLACK_TITLE: "Success"
          SLACK_MESSAGE: "All checks passed :white_check_mark:"
          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }} 

This job now notified us whenever all the CI checks passed. The remaining part for this section was notification if any one of the 3 CI checks fails. We used if syntax from the workflow to achieve this action. The if statement has higher precedence than needs statement. The slack_failure job is made dependent on the slack_success job. And, the slack_failure job will only started if the slack_success job had failed. The failure of a job or previous step action can be checked using statements failed() or cancelled() expression.

slack_failure:
    name: slack
    runs-on: ubuntu-latest
    needs: [slack_success]
    if: failure() || cancelled()
    steps:
      - uses: actions/checkout@v1
      - name: Success Notify
        uses: rtCamp/action-slack-notify@master
        env:
          SLACK_USERNAME: "Action Bot"
          SLACK_ICON: "https://github.com/freeletics.png"
          SLACK_COLOR: "#FF0000"
          SLACK_TITLE: "Failure"
          SLACK_MESSAGE: "Some checks failed :x:"
          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }} 

Outro

The complete workflow code for the CI is shown at the bottom of post

With this we could replicate our current CI implementation completely using Github Actions. However, we haven’t yet started using it for our daily development. Some of the reasons are:

1. The support for latest ruby is late. We are already using ruby 2.6.5 for our apps, but actions/setup-ruby is yet to support it. There has been discussion about open sourcing the process of adding newer versions of ruby. Comment.
2. The gems are not cached between builds. For every build, bundler needs to install the gems fresh, and it takes quite a time. There has also been discussion about it and confirmation that Github Actions is working on adding caching. Comment

References:

1. Workflow syntax for GitHub Actions
2. Contexts and expression syntax for GitHub

Complete Workflow Yaml

name: CI
on: [push, pull_request]
jobs:
  specs:
    name: specs
    runs-on: ubuntu-latest
    services:
      postgres:
        image: postgres:10.8
        ports:
        - 5432:5432
        env:
          POSTGRES_PASSWORD: ""
    steps:
    - name: Install libraries
      run: |
        sudo apt-get update
        sudo apt-get install -y postgresql-client libpq-dev

    - name: Configure databases
      run: |
        echo "Postgres"
        psql -h localhost -c 'create database "test-database";' -U postgres

    - uses: actions/checkout@v1

    - name: Set up Ruby
      uses: actions/setup-ruby@v1
      with:
        ruby-version: 2.6.5

    - name: Install bundler and gems
      run: |
        gem install bundler --no-document
        bundle config GITHUB__COM $GITHUB_ACCESS_TOKEN
        bundle install --jobs 4 --retry 3
      env:
        GITHUB_ACCESS_TOKEN: $

    - name: Run tests
      run: |
        pg_config --version
        bin/rails db:schema:load RAILS_ENV=test
        bin/rspec
  rubocop:
    name: rubocop
    runs-on: ubuntu-latest
    steps:
      - name: Install libraries
        run: |
          sudo apt-get update
          sudo apt-get install -y libpq-dev
      - name: Set up Ruby
        uses: actions/setup-ruby@v1
        with:
          ruby-version: 2.6.5

      - uses: actions/checkout@v1

      - name: Install bundler and rubocop
        run: |
          gem install bundler --no-document
          bundle config GITHUB__COM $GITHUB_ACCESS_TOKEN
          bundle install --jobs 4 --retry 3 --with=test
        env:
          GITHUB_ACCESS_TOKEN: $

      - name: Run rubocop checks
        run: bundle exec rubocop -D -c .rubocop.yml

  swagger:
    name: swagger
    runs-on: ubuntu-latest
    steps:
      - name: Install Node
        uses: actions/setup-node@v1
        with:
          node-version: 10.9.0

      - name: Install swagger cli
        run: |
          npm install -g swagger-cli

      - uses: actions/checkout@v1

      - name: Run swagger linter check
        run: |
          for f in doc/freeletics_api_v{1,2,3}.yml; do swagger-cli validate $f || break 0; done

  slack_success:
    name: slack
    runs-on: ubuntu-latest
    needs: [swagger, rubocop, specs]
    steps:
      - uses: actions/checkout@v1
      - name: Success Notify
        uses: rtCamp/action-slack-notify@master
        env:
          SLACK_USERNAME: "Action Bot"
          SLACK_ICON: "https://github.com/freeletics.png"
          SLACK_TITLE: "Success"
          SLACK_MESSAGE: "All checks passed :white_check_mark:"
          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }} 

  slack_failure:
    name: slack
    runs-on: ubuntu-latest
    needs: [slack_success]
    if: failure() || cancelled()
    steps:
      - uses: actions/checkout@v1
      - name: Success Notify
        uses: rtCamp/action-slack-notify@master
        env:
          SLACK_USERNAME: "Action Bot"
          SLACK_ICON: "https://github.com/freeletics.png"
          SLACK_COLOR: "#FF0000"
          SLACK_TITLE: "Failure"
          SLACK_MESSAGE: "Some checks failed :x:"
          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}