GitLab QA orchestrator

Definitions

GitLab QA framework: A framework that allows developers to write end-to-end tests simply and efficiently. Located at gitlab-org/gitlab-ce@qa/qa/.
GitLab QA instance-level scenarios: RSpec scenarios that use the GitLab QA framework and Capybara to setup and perform individual end-to-end tests against a live GitLab instance. Located at gitlab-org/gitlab-ce@qa/qa/specs/features/.
GitLab QA orchestrator (this project): An orchestration tool that enables running various QA test suites in a simple manner.
GitLab QA orchestrated scenarios: Scenarios where containers are started, configured, and execute instance-level scenarios against a running GitLab instance. Located at gitlab-org/gitlab-qa@lib/gitlab/qa/scenario/test/.

Goals and objectives

GitLab consists of multiple pieces configured and packaged by GitLab Omnibus.

The purpose of the QA end-to-end test suite is to verify that all pieces integrate well together.

Testing changes in merge requests before the merge

The ultimate goal is to make it possible to run the QA test suite for any merge request, even before merging code into the master branch.

We can run tests against any instance of GitLab

GitLab QA is a click-driven, black-box testing tool. We also use it to run tests against the staging environment, and we strive to make it useful for our users as well.

GitLab QA tests running in the CI/CD environment

Manual steps should not be needed to run the QA test suite. GitLab QA orchestrator is CI/CD environment native, which means that we should add new features and tests when we are comfortable with running new code in the CI/CD environment.

GitLab QA test failures are reproducible locally

Despite the fact that GitLab QA orchestrator has been built to run in the CI/CD environment, it is really important to make it easy for developers to reproduce test failures locally. It is much easier to debug things locally, than in the CI/CD environment.

To make it easier to reproduce test failures locally we have published the gitlab-qa gem on rubygems.org and we are using exactly the same approach to run tests in the CI/CD environment.

It means that using the gitlab-qa CLI tool, which orchestrates the test environment and runs the GitLab QA test suite, is a reproducible way of running tests locally and in the CI/CD environment.

It also means that we cannot have custom code in .gitlab-ci.yml to, for example, start new containers / services.

Test the installation / deployment process too

We distribute GitLab in a package (like a Debian package or a Docker image) so we want to test the installation process to ensure that our package is not broken.

But we are also working on making GitLab be a cloud native product. This means that, for example, using Helm becomes yet another installation / deployment process that we want to test with GitLab QA.

Considering our goal of being able to test all changes in merge requests, it is especially important to be able to test our Kubernetes deployments, as that is essential to scaling our test environments to efficiently handle a large number of tests.

Documentation

How do we use it

Currently, we execute the test suite against GitLab Docker images created by Omnibus nightly via a pipeline in the nightly project.

We also execute the test suite nightly against our staging environment via a pipeline in the staging project.

Finally, we trigger GitLab QA pipelines whenever someone clicks package-and-qa manual action in a merge request.

How can you use it

The GitLab QA tool is published as a Ruby Gem.

You can install it with gem install gitlab-qa. It will expose a gitlab-qa command in your system.

If you want to run the scenarios against your GDK and/or develop them on Mac OS, please read Run QA tests against your GDK setup as there are caveats and things that may work differently.

All the scenarios you can run are described in the What tests can be run? documentation.

Note: The GitLab QA tool requires that Docker is installed.

Command-line options

In addition to the arguments you can use to specify the scenario and tests to run, you can use the following options to control the tool's behavior.

Note: These are gitlab-qa options so if you specify RSpec options as well, including test file paths, be sure to add these options before the -- that indicates that subsequent arguments are intended for RSpec.

Option	Description
`--no-teardown`	Skip teardown of containers after the scenario completes
`--no-tests`	Orchestrates the docker containers but does not run the tests. Implies `--no-teardown`

For example, the following command would start an EE GitLab Docker container and would leave the instance running, but would not run the tests:

$ gitlab-qa Test::Instance::Image EE --no-tests

GitLab QA will have automatically run the docker ps command to show the port that container is running on, for example:

...
Skipping tests.
The orchestrated docker containers have not been removed.
Docker shell command: `docker ps`
CONTAINER ID  IMAGE                     ... PORTS
fdeffd791b69  gitlab/gitlab-ee:nightly      22/tcp, 443/tcp, 0.0.0.0:32768->80/tcp

You could then run tests against that instance in a similar way to running tests against GDK. This can be useful if you want to run and debug a specific test, for example:

# From /path/to/gdk/gitlab/qa
$ bundle exec bin/qa Test::Instance::All http://localhost:32768 -- qa/specs/features/browser_ui/3_create/merge_request/create_merge_request_spec.rb

How to add new tests

Please see the Beginner's guide to writing end-to-end tests.

Test cases and scripts to run them are located in the GitLab FOSS and GitLab repositories under the qa/ directory, so please also check the documentation there.

Contributing

Please see the contribution guidelines.