End-to-end Testing contribute

What is end-to-end testing?

End-to-end (e2e) testing is a strategy used to check whether your application works as expected across the entire software stack and architecture, including integration of all micro-services and components that are supposed to work together.

How do we test GitLab?

To test GitLab, we:

  1. Use CNG to build GitLab Cloud Native packages.
  2. Deploy these packages using the gitlab-cng orchestrator CLI tool to create a running instance of GitLab to run E2E tests against.

Additionally, we use the GitLab Development Kit (GDK) as a test environment that can be deployed quickly for faster test feedback.

Testing nightly builds

We run scheduled pipelines each night to test nightly builds created by Omnibus. You can find these pipelines at https://gitlab.com/gitlab-org/gitlab/-/pipeline_schedules (requires the Developer role). Results are reported in the #e2e-run-master Slack channel.

Testing staging

We run scheduled pipelines each night to test staging. You can find these pipelines at https://gitlab.com/gitlab-org/quality/staging/pipelines (requires the Developer role). Results are reported in the #e2e-run-staging Slack channel.

Testing code in merge requests

End-to-end test pipelines describes pipeline setup responsible for running E2E testing within merge requests.

Using the test-on-omnibus job

It is possible to run end-to-end tests for a merge request by triggering the e2e:test-on-omnibus manual action in the qa stage (not available for forks).

This runs end-to-end tests against a custom EE (with an Ultimate license) Docker image built from your merge request’s changes.

Manual action that starts end-to-end tests is also available in gitlab-org/omnibus-gitlab merge requests.

With merged results pipelines

In a merged results pipeline, the pipeline runs on a new ref that contains the merge result of the source and target branch.

The end-to-end tests on a merged results pipeline would use the new ref instead of the head of the merge request source branch.

x1y1z1 - master HEAD
d1e1f1 - merged results (CI_COMMIT_SHA)
Merged results pipeline
E2E tests
Running custom tests

The existing scenarios that run in the downstream gitlab-qa-mirror pipeline include many tests, but there are times when you might want to run a test or a group of tests that are different than the groups in any of the existing scenarios.

For example, when we dequarantine a flaky test we first want to make sure that it’s no longer flaky. We can do that by running _ee:quarantine manual job. When selecting the name (not the play icon) of manual job, you are prompted to enter variables. You can use any of the variables that can be used with gitlab-qa as well as these:

VariableDescription
QA_SCENARIOThe scenario to run (default Test::Instance::Image)
QA_TESTSThe tests to run (no default, which means run all the tests in the scenario). Use file paths as you would when running tests by using RSpec, for example, qa/specs/features/ee/browser_ui would include all the EE UI tests.
QA_RSPEC_TAGSThe RSpec tags to add (default --tag quarantine)

For now, manual jobs with custom variables don’t use the same variable when retried, so if you want to run the same tests multiple times, specify the same variables in each custom-parallel job (up to as many of the 10 available jobs that you want to run).

Selective test execution

In order to limit amount of tests executed in a merge request, dynamic selection of which tests to execute is present. Algorithm of which tests to run is based on changed files and merge request labels. Following criteria determine which tests will run:

  1. Changes in qa framework code would execute the full suite
  2. Changes in particular _spec.rb file in qa folder would execute only that particular test. In this case knapsack will not be used to run jobs in parallel.
Selective test execution based on code path mappings

Mapping based selective test execution is currently in use for test-on-gdk pipeline. For more information, see epic 47.

Overriding selective test execution

To override selective test execution and trigger the full suite, label pipeline:run-all-e2e should be added to particular merge request.

Skipping end-to-end tests

In some cases, it may not be necessary to run the end-to-end test suite.

Examples could include:

  • ~“Stuff that should Just Work”
  • Small refactors
  • A small requested change during review, that doesn’t warrant running the entire suite a second time

Skip running end-to-end tests by applying the pipeline:skip-e2e label to the merge request.

There is a risk in skipping end-to-end tests. Use caution and discretion when applying this label. The end-to-end test suite is the last line of defense before changes are merged into the default branch. Skipping these tests increases the risk of introducing regressions into the codebase.

Dynamic parallel job scaling

To maintain consistent pipeline run times, the CI/CD job count for each particular E2E test suite is scaled dynamically based on total run time of tests in the suite. The generate_e2e_pipelines Rake task creates CI/CD YAML files that:

  • Create the correct number of parallel jobs.
  • Skip particular jobs entirely if no tests would be executed.

This functionality works in tandem with selective test execution to optimize pipeline run time and costs as much as possible based on particular changes within merge request.

Design outline

Dynamic job scaling relies on Test Scenario classes. This abstraction encapsulates the following:

  • The RSpec tags a particular scenario should execute.
  • An optional spec file pattern that can limit the scenario to particular spec files.
  • A pipeline mapping which defines the pipeline types and jobs a particular scenario would run in merge request pipelines.

The PipelineCreator class generates pipeline YAML files with dynamically adjusted parallel job counts. Before pipeline YAML generation, PipelineCreator iterates over all defined Test Scenario classes and creates a mapping which contains a total of calculated test run time for each CI/CD job. Based on this information, the pipeline YAML file is generated with correctly adjusted parallel job count.

PipelineCreator additionally takes input from selective test execution to further reduce the total number of tests that would be executed.

For an example of how to create a new scenario that would run this scenario in merge request pipelines and scale parallel jobs dynamically, see Adding new jobs to E2E test pipelines.

Test pipeline tools and configuration

Test parallelization

Our CI setup uses the knapsack gem to enable test parallelization. Knapsack reports are automatically generated and stored in the knapsack-reports GCS bucket within the gitlab-qa-resources project. The KnapsackReport helper manages the report generation and upload process.

Test metrics

To enhance test health visibility, a custom setup exports the pipeline’s test execution results to an InfluxDB instance, with results visualized on Grafana dashboards.

Test reports

Allure report

For additional test results visibility, tests that run on pipelines generate and host Allure test reports.

The QA framework is using the Allure RSpec gem to generate source files for the Allure test report. An additional job in the pipeline:

  • Fetches these source files from all test jobs.
  • Generates and uploads the report to the S3 bucket gitlab-qa-allure-report located in AWS group project eng-quality-ops-ci-cd-shared-infra.

A common CI template for report uploading is stored in allure-report.yml.

Merge requests

When these tests are executed in the scope of merge requests, the Allure report is uploaded to the GCS bucket and a bot comment is added linking to their respective reports.

Scheduled pipelines

Scheduled pipelines for these tests contain a generate-allure-report job under the Report stage. They also output a link to the current test report. Each type of scheduled pipeline generates a static link for the latest test report according to its stage. You can find a list of this in the GitLab handbook.

Provisioning

Provisioning of all components is performed by the engineering-productivity-infrastructure project.

Exporting metrics in CI

Use these environment variables to configure metrics export:

VariableRequiredInformation
QA_INFLUXDB_URLtrueShould be set to https://influxdb.quality.gitlab.net. No default value.
QA_INFLUXDB_TOKENtrueInfluxDB write token that can be found under Influxdb auth tokens document in Gitlab-QA 1Password vault. No default value.
QA_RUN_TYPEfalseArbitrary name for test execution, like e2e:test-on-omnibus. Automatically inferred from the project name for live environment test executions. No default value.
QA_EXPORT_TEST_METRICSfalseFlag to enable or disable metrics export to InfluxDB. Defaults to false.
QA_SAVE_TEST_METRICSfalseFlag to enable or disable saving metrics as JSON file. Defaults to false.

How do you run the tests?

If you are not testing code in a merge request, there are two main options for running the tests. If you want to run the existing tests against a live GitLab instance or against a pre-built Docker image, use the GitLab QA orchestrator. See also examples of the test scenarios you can run by using the orchestrator.

On the other hand, if you would like to run against a local development GitLab environment, you can use the GitLab Development Kit (GDK). Refer to the instructions in the QA README and the section below.

Running tests that require special setup

Learn how to perform tests that require special setup or consideration to run on your local environment.

How do you write tests?

Before you write new tests, review the GitLab QA architecture.

After you’ve decided where to put test environment orchestration scenarios and instance-level scenarios, take a look at the GitLab QA README, the GitLab QA orchestrator README, and the already existing instance-level scenarios.

Consider not writing an end-to-end test

We should follow these best practices for end-to-end tests:

  • Do not write an end-to-end test if a lower-level feature test exists. End-to-end tests require more work and resources.
  • Troubleshooting for end-to-end tests can be more complex as connections to the application under test are not known.

Continued reading

Getting started with E2E testing

  • Beginner’s Guide: An introductory guide to help new contributors get started with E2E testing
    • Flows: Overview of Flows used to capture reusable sequences of actions in tests
    • Page objects: Explanation of page objects and their role in test design
    • Resources: Overview of Resources class that used for creating test data

Best practices

Testing infrastructure

Running and troubleshooting tests

Miscellaneous

Where can you ask for help?

You can ask question in the #test-platform channel on Slack (GitLab internal) or you can find an issue you would like to work on in the gitlab issue tracker.