CICD Demo Project

Description

This project demonstrates best-practices of a CICD pipeline development. I've used Github Actions, (self-hosted) Runners, Browserstack integration, Junit unit test cases, Playwright browser test cases and the creation of a Docker image.

Environment

• Local LAN comprising:
  • 3x self-hosted runners: 2x openSUSE Leap 15.5 (CLI & GNOME), 1x Windows 10 Pro
• IDE Intellij 2024.1 CE with Codiumate AI plugin
• OpenJDK 17 | 19, "temurin" or "zulu" on self-hosted runners
• Github Free account
• Browserstack trial account (supports Windows Chrome)
• Playwright 1.42
• Junit5
• Allure 2.13 & custom reporting classes
• DockerDesktop 4.29

CICD Pipeline Strategy

1. After a Push or Pull Request into master, CICD_Main_Pipeline.yml is executed. This workflow takes a build through three test levels to a fictional production environment
2. CICD_Main_Pipeline.yml performs the following steps:
   • Setup: JDK 19 (OpenJDK temurin|zulu distribution) is deployed to the runner (self-hosted). THe latest version of the this script:
     • Checks if the correct version of JDK is already installed and if it is, proceeds to build
     • If not, the cache is checked and if present downloaded from there
     • If not, the JDK is retrieved from a prebuilt image provided by GitHub.
     • Note I am not convinced this logic is required as the JDK will not change much over the lifespan of a single project but it is included as POC
   • Build: Maven builds the project and runs tests. Maven must be installed on all Runners beforehand.
   • Tests are Junit tests with and without Playwright running on local or Browserstack browsers
   • Docker: Docker creates a new Docker image from the project Dockerfile, copies and executes the mainClass from project Jar (no test execution)
   • My workflow yml uses various strategies to test on OS, browsers and platforms. OS and browsers are configured via the strategy syntax and platfoms via Maven profiles
   • Test Reporting. By default, Junit provides limited test reporting capability. Certainly, for client projects in regulated markets more transparency is necessary. The addition of
3. Custom reporting classes (DisplayNameReporter.class) or reporting frameworks, e.g. Allure, improve this greatly.

Workflow Schedule

here

Note:

Caching is provided by actions/cache but, whilst useful, caches in the cloud. For project dependencies there is a valid usecase to cache in the cloud. OpenJDK packages have proved very slow to sync onto a self-hosted runner and given the lack of need to constantly update, I have added the follow step to my workflow which checks for a local installation of (the correct version) of Java (bashrc version):

  - name: Check Java Version
    id: check-java
    run: |
      JAVA_VERSION=19 # change as needed
      if type -p java; then
        echo "Java found in PATH"
        _java=java
      elif [[ -n "$JAVA_HOME" ]] && [[ -x "$JAVA_HOME/bin/java" ]]; then
        echo "Java found in JAVA_HOME"
        _java="$JAVA_HOME/bin/java"
      else
        echo "Java not found"
        echo "java_installed=false" >> $GITHUB_ENV
        exit 1
      fi
      
      if [[ "$_java" ]]; then
        version=$("$_java" -version 2>&1 | awk -F[\"_] '{print $2}')
        echo "Java version: $version"
        if [[ "$version" == "$JAVA_VERSION"* ]]; then
          echo "Java version $JAVA_VERSION is already installed."
          echo "java_installed=true" >> $GITHUB_ENV
        else
          echo "Installed Java version does not match required version."
          echo "java_installed=false" >> $GITHUB_ENV
          exit 1
        fi
      fi

  - name: Set up cache
    if: env.java_installed != 'true'
    uses: actions/[email protected]
    with:
      key: ${{ runner.os }}-java-${{ hashFiles('**/build.gradle') }}
      restore-keys: |
        ${{ runner.os }}-java-

Limitations

Listed here