Home Original page

GitHub - ppodgorsek/docker-robot-framework: Robot Framework in Docker

Robot Framework in Docker, with Firefox, Chrome and Microsoft Edge

Table of contents

What is it?

This project consists of a container image containing a Robot Framework installation.

This installation also contains Firefox, Chrome, Microsoft Edge, along with the Selenium and Playwright/RFBrowser library for Robot Framework.

Versioning

The versioning of this image follows the one of Robot Framework:

  • Major version matches the one of Robot Framework
  • Minor and patch versions are specific to this project (allows to update the versions of the other dependencies)

The versions used are:

As stated by the official GitHub project, starting from version 3.0, Selenium2Library is renamed to SeleniumLibrary and this project exists mainly to help with transitioning. The Selenium2Library 3.0.0 is also the last release and for new releases, please look at the SeleniumLibrary project.

Running the container

This container can be run using the following command:

docker run \
    -v <local path to the reports' folder>:/opt/robotframework/reports:Z \
    -v <local path to the test suites' folder>:/opt/robotframework/tests:Z \
    ppodgorsek/robot-framework:<version>

Switching browsers

Browsers can be easily switched. It is recommended to define ${BROWSER} %{BROWSER} in your Robot variables and to use ${BROWSER} in your test cases. This allows to set the browser in a single place if needed.

When running your tests, simply add -e BROWSER=chrome, -e BROWSER=firefox or -e BROWSER=edgeto the run command.

Please note: edge will work with Selenium but not the Browser Library, as the latter currently doesn't have an easy mechanism to install additional browsers. Playwright, on which the Browser library relies, cannot install additional browsers on Linux platforms other than Ubuntu/Debian and suggests using Chromium to test Microsoft Edge scenarios, unless you require Edge-specific capabilities.

Changing the container's screen resolution

It is possible to define the settings of the virtual screen in which the browser is run by changing several environment variables:

  • SCREEN_COLOUR_DEPTH (default: 24)
  • SCREEN_HEIGHT (default: 1080)
  • SCREEN_WIDTH (default: 1920)

Changing the container's tests and reports directories

It is possible to use different directories to read tests from and to generate reports to. This is useful when using a complex test file structure. To change the defaults, set the following environment variables:

  • ROBOT_REPORTS_DIR (default: /opt/robotframework/reports)
  • ROBOT_TESTS_DIR (default: /opt/robotframework/tests)

Parallelisation

It is possible to parallelise the execution of your test suites. Simply define the ROBOT_THREADS environment variable, for example:

docker run \
    -e ROBOT_THREADS=4 \
    ppodgorsek/robot-framework:latest

By default, there is no parallelisation.

Parallelisation options

When using parallelisation, it is possible to pass additional pabot options, such as --testlevelsplit, --argumentfile, --ordering, etc. These can be passed by using the PABOT_OPTIONS environment variable, for example:

docker run \
    -e ROBOT_THREADS=4 \
    -e PABOT_OPTIONS="--testlevelsplit" \
    ppodgorsek/robot-framework:latest

Passing additional options

RobotFramework supports many options such as --exclude, --variable, --loglevel, etc. These can be passed by using the ROBOT_OPTIONS environment variable, for example:

docker run \
    -e ROBOT_OPTIONS="--loglevel DEBUG" \
    ppodgorsek/robot-framework:latest

Testing emails

This project includes the IMAP library which allows Robot Framework to connect to email servers.

A suggestion to automate email testing is to run a Mailcatcher instance in Docker which allows IMAP connections. This will ensure emails are discarded once the tests have been run.

Dealing with Datetimes and Timezones

This project is meant to allow your tests to run anywhere. Sometimes that can be in a different timezone than your local one or of the location under test. To help solve such issues, this image includes the DateTimeTZ Library.

To set the timezone used inside the Docker image, you can set the TZ environment variable:

docker run \
    -e TZ=America/New_York \
    ppodgorsek/robot-framework:latest

Installing additional dependencies

It is possible to install additional dependencies dynamically at runtime rather than having to extend this image.

To do so, simply mount a text file containing the list of dependencies you would like to install using pip: (by default, this file is empty if not mounted)

docker run \
    -v <local path to the dependency file>:/opt/robotframework/pip-requirements.txt:Z \
    -v <local path to the test suites' folder>:/opt/robotframework/tests:Z \
    ppodgorsek/robot-framework:latest

The file must follow Pip's official requirements file format.

Here is a example of what such a file could contain:

robotframework-docker==1.4.2
rpa==1.50.0

For large dependencies, it is still recommended to extend the project's image and to add them there, to avoid delaying the CI/CD pipelines with repeated dependency installations.

Rerunning tests

Failing tests can be rerun by setting the environment variable ROBOT_RERUN_MAX_ROUNDS to a value above 0. All reruns of failed tests are executed without parallelization. The number in environment variable ROBOT_RERUN_MAX_ROUNDS dictates how many rerun-rounds are made at maximum. All rerun-rounds will only test what failed in the previous round. The report files combine the results of all rounds, the last round providing the final result.

The default value for ROBOT_RERUN_MAX_ROUNDS is 0, meaning that tests will not be executed again if they fail.

docker run \
    -e ROBOT_RERUN_MAX_ROUNDS=1 \
    ppodgorsek/robot-framework:latest

Security consideration

By default, containers are implicitly run using --user=1000:1000, please remember to adjust that command-line setting accordingly, for example:

docker run \
    --user=1001:1001 \
    ppodgorsek/robot-framework:latest

Remember that that UID/GID should be allowed to access the mounted volumes in order to read the test suites and to write the output.

Additionally, it is possible to rely on user namespaces to further secure the execution. This is well described in the official container documentation:

This is a good security practice to make sure containers cannot perform unwanted changes on the host. In that sense, Podman is probably well ahead of Docker by not relying on a root daemon to run its containers.

Continuous integration

Azure DevOps pipeline

Azure DevOps provides a Docker@2 task but it only allows a limited number of commands, it is therefore required to use a script instead:

- job: FunctionalRegressionTests
  displayName: Robot Framework tests
  steps:
    - script: |
        set -x

        mkdir -p $(Build.Repository.LocalPath)/robot-framework-reports
        chmod -R ugo+rwx $(Build.Repository.LocalPath)/robot-framework-reports

        # The ROBOT_OPTIONS allow to generate a specific output file that can be
        # used to publish test results in the Azure DevOps pipeline run
        # Remember to replace ${{ parameters.robotFrameworkFolder }} by the correct
        # folder name in your repository
        docker run --rm \
          --shm-size=1g \
          -v $(Build.Repository.LocalPath)/robot-framework-reports:/opt/robotframework/reports:Z \
          -v $(Build.Repository.LocalPath)/${{ parameters.robotFrameworkFolder }}:/opt/robotframework/tests:Z \
          -e ROBOT_OPTIONS="-x outputxunit.xml" \
          -t docker.io/ppodgorsek/robot-framework:latest
      displayName: Run Robot Framework tests

    - task: PublishTestResults@2
      condition: always()
      displayName: Publish Robot Framework test results
        inputs:
          testResultsFormat: 'JUnit'
          testResultsFiles: '**/outputxunit.xml'
          searchFolder: '$(Build.Repository.LocalPath)/robot-framework-reports'
          failTaskOnFailedTests: true
          failTaskOnFailureToPublishResults: true
          failTaskOnMissingResultsFile: true

Jenkins pipeline

It is possible to run the project from within a Jenkins pipeline by relying on the shell command line directly:

pipeline {
  agent any
  stages {
    stage('Functional regression tests') {
      steps {
        sh "docker run --shm-size=1g -e BROWSER=firefox -v $WORKSPACE/robot-tests:/opt/robotframework/tests:Z -v $WORKSPACE/robot-reports:/opt/robotframework/reports:Z ppodgorsek/robot-framework:latest"
      }
    }
  }
}

The pipeline stage can also rely on a Docker agent, as shown in the example below:

pipeline {
  agent none
  stages {
    stage('Functional regression tests') {
      agent { docker {
        image 'ppodgorsek/robot-framework:latest'
        args '--shm-size=1g -u root' }
      }
      environment {
        BROWSER = 'firefox'
        ROBOT_TESTS_DIR = "$WORKSPACE/robot-tests"
        ROBOT_REPORTS_DIR = "$WORKSPACE/robot-reports"
      }
      steps {
        sh '''
          /opt/robotframework/bin/run-tests-in-virtual-screen.sh
        '''
      }
    }
  }
}

Defining a test run ID

When relying on Continuous Integration tools, it can be useful to define a test run ID such as the build number or branch name to avoid overwriting consecutive execution reports.

For that purpose, the ROBOT_TEST_RUN_ID variable was introduced:

  • If the test run ID is empty, the reports folder will be: ${ROBOT_REPORTS_DIR}/
  • If the test run ID was provided, the reports folder will be: ${ROBOT_REPORTS_DIR}/${ROBOT_TEST_RUN_ID}/

It can simply be passed during the execution, such as:

docker run \
    -e ROBOT_TEST_RUN_ID="feature/branch-name" \
    ppodgorsek/robot-framework:latest

By default, the test run ID is empty.

Upload test reports to an AWS S3 bucket

To upload the report of a test run to an S3 bucket, you need to define the following environment variables:

docker run \
    -e AWS_ACCESS_KEY_ID=<your AWS key> \
    -e AWS_SECRET_ACCESS_KEY=<your AWS secret> \
    -e AWS_DEFAULT_REGION=<your AWS region e.g. eu-central-1> \
    -e AWS_BUCKET_NAME=<name of your S3 bucket> \
    ppodgorsek/robot-framework:latest

Testing this project

Not convinced yet? Simple tests have been prepared in the test/ folder, you can run them using the following commands:

# Using Chromium
docker run \
    -v `pwd`/reports:/opt/robotframework/reports:Z \
    -v `pwd`/test:/opt/robotframework/tests:Z \
    -e BROWSER=chrome \
    ppodgorsek/robot-framework:latest

# Using Firefox
docker run \
    -v `pwd`/reports:/opt/robotframework/reports:Z \
    -v `pwd`/test:/opt/robotframework/tests:Z \
    -e BROWSER=firefox \
    ppodgorsek/robot-framework:latest

For Windows users who use PowerShell, the commands are slightly different:

# Using Chromium
docker run \
    -v ${PWD}/reports:/opt/robotframework/reports:Z \
    -v ${PWD}/test:/opt/robotframework/tests:Z \
    -e BROWSER=chrome \
    ppodgorsek/robot-framework:latest

# Using Firefox
docker run \
    -v ${PWD}/reports:/opt/robotframework/reports:Z \
    -v ${PWD}/test:/opt/robotframework/tests:Z \
    -e BROWSER=firefox \
    ppodgorsek/robot-framework:latest

Screenshots of the results will be available in the reports/ folder.

Troubleshooting

Chromium is crashing

Chrome drivers might crash due to the small size of /dev/shm in the docker container:

UnknownError: session deleted because of page crash

This is a known bug of Chromium.

To avoid this error, please change the shm size when starting the container by adding the following parameter: --shm-size=1g (or any other size more suited to your tests)

Accessing the logs

In case further investigation is required, the logs can be accessed by mounting their folder. Simply add the following parameter to your run command:

  • Linux/Mac: -v `pwd`/logs:/var/log:Z
  • Windows: -v ${PWD}/logs:/var/log:Z

Chromium allows to set additional environment properties, which can be useful when debugging:

  • webdriver.chrome.verboseLogging=true: enables the verbose logging mode
  • webdriver.chrome.logfile=/path/to/chromedriver.log: sets the path to Chromium's log file

Error: Suite contains no tests

When running tests, an unexpected error sometimes occurs:

[Error] Suite contains no tests.

There are two main causes to this:

  • Either the test folder is not the right one,
  • Or the permissions on the test folder/test files are too restrictive.

As there can sometimes be issues as to where the tests are run from, make sure the correct folder is used by trying the following actions:

  • Use a full path to the folder instead of a relative one,
  • Replace any`pwd`or ${PWD} by the full path to the folder.

It is also important to check if Robot Framework is allowed to access the resources it needs, i.e.:

  • The folder where the tests are located,
  • The test files themselves.

Database tests are failing in spite of the DatabaseLibrary being present

As per their official project page, the Robot Framework DatabaseLibrary contains utilities meant for Robot Framework's usage. This can allow you to query your database after an action has been made to verify the results. This is compatible with any Database API Specification 2.0 module.

It is anyway mandatory to extend the container image to install the specific database module relevant to your tests, such as:

Supported devices and architectures

As mentioned on the Docker Hub, the project has been built and uploaded for linux/amd64 and linux/arm64 architectures only.

As mentioned in the official documentation, Podman and Docker provide a --platform option which selects a given application architecture, such as:

docker run \
    --platform linux/amd64 \
    -v <local path to the reports' folder>:/opt/robotframework/reports:Z \
    -v <local path to the test suites' folder>:/opt/robotframework/tests:Z \
    ppodgorsek/robot-framework:<version>

It is to be noted that Microsoft Edge is not available for Linux ARM. Please visit the official Microsoft Edge website and the Arm developer website for more information.

If this does not solve your platform-related issues, you might have to rebuild the image for your device/platform, specifying that --platform option during the build and run.

Container image is too large

Unfortunately, the image size is due to the presence of both Selenium and Playwright. (the latter requiring its own set of browsers)

Should you need to use Robot Framework in embedded devices, this is too much.

Instead, a basic multi-arch container (under 200MB) can be found in this public container registry: Robot Framework for Embedded Multi-Arch Container.

It is built as part of the Robot Framework for Embedded.

Please contribute!

Have you found an issue? Do you have an idea for an improvement? Feel free to contribute by submitting it on the GitHub project.