e2e:package-and-test

The e2e:package-and-test child pipeline is the main executor of E2E testing for the GitLab platform. The pipeline definition has several dynamic components to reduce the number of tests being executed in merge request pipelines.

Setup

Pipeline setup consists of:

The e2e-test-pipeline-generate job in the prepare stage of the main GitLab pipeline.
The e2e:package-and-test job in the qa stage, which triggers the child pipeline that is responsible for building the omnibus package and running E2E tests.

e2e-test-pipeline-generate

This job consists of two components that implement selective test execution:

The detect_changes Rake task determines which e2e specs should be executed in a particular merge request pipeline. This task analyzes changes in a particular merge request and determines which specs must be executed. Based on that, a dry-run of every scenario executes and determines if a scenario contains any executable tests. Selective test execution uses these criteria to determine which specific tests to execute.
generate-e2e-pipeline is executed, which generates a child pipeline YAML definition file with appropriate environment variables.

e2e:package-and-test

E2E test execution pipeline consists of several stages which all support execution of E2E tests.

.pre

This stage is responsible for the following tasks:

Fetching knapsack reports that support parallel test execution.
Triggering downstream pipeline which builds the omnibus-gitlab Docker image.

test

This stage runs e2e tests against different types of GitLab configurations. The number of jobs executed is determined dynamically by e2e-test-pipeline-generate job.

report

This stage is responsible for allure test report generation.

Adding new jobs

Selective test execution depends on a set of rules present in every job definition. A typical job contains the following attributes:

variables:
  QA_SCENARIO: Test::Integration::MyNewJob
rules:
  - !reference [.rules:test:qa, rules]
  - if: $QA_SUITES =~ /Test::Integration::MyNewJob/
  - !reference [.rules:test:manual, rules]

In this example:

QA_SCENARIO: Test::Integration::MyNewJob: name of the scenario class that is passed to the gitlab-qa executor.
!reference [.rules:test:qa, rules]: main rule definition that is matched for pipelines that should execute all tests. For example, when changes to qa framework are present.
if: $QA_SUITES =~ /Test::Integration::MyNewJob/: main rule responsible for selective test execution. QA_SUITE is the name of the scenario abstraction located in qa framework.

QA_SUITE is not the same as QA_SCENARIO, which is passed to the gitlab-qa executor. For consistency, it usually has the same name. QA_SUITE abstraction class usually contains information on what tags to run and optionally some additional setup steps.
!reference [.rules:test:manual, rules]: final rule that is always matched and sets the job to manual so it can still be executed on demand, even if not set to execute by selective test execution.

Considering example above, perform the following steps to create a new job:

Create new scenario type my_new_job.rb in the integration directory of the gitlab-qa project and release new version so it's generally available.

Create new scenario my_new_job.rb in integration directory of the qa framework. In the most simple case, this scenario would define RSpec tags that should be executed:

module QA
  module Scenario
    module Test
      module Integration
        class MyNewJob < Test::Instance::All
          tags :some_special_tag
        end
      end
    end
  end
end

Add new job definition in the main.gitlab-ci.yml pipeline definition:

ee:my-new-job:
  extends: .qa
  variables:
    QA_SCENARIO: Test::Integration::MyNewJob
  rules:
    - !reference [.rules:test:qa, rules]
    - if: $QA_SUITES =~ /Test::Integration::MyNewJob/
    - !reference [.rules:test:manual, rules]

Parallel jobs

For selective execution to work correctly with job types that require running multiple parallel jobs, a job definition typically must be split into parallel and selective variants. Splitting is necessary so that when selective execution executes only a single spec, multiple unnecessary jobs are not spawned. For example:

ee:my-new-job-selective:
  extends: .qa
  variables:
    QA_SCENARIO: Test::Integration::MyNewJob
  rules:
    - !reference [.rules:test:qa-selective, rules]
    - if: $QA_SUITES =~ /Test::Integration::MyNewJob/
ee:my-new-job:
  extends:
    - .parallel
    - ee:my-new-job-selective
  rules:
    - !reference [.rules:test:qa-parallel, rules]
    - if: $QA_SUITES =~ /Test::Integration::MyNewJob/