GitLab Data Seeder

GitLab Data Seeder (GDS) is a test data seeding harness, that can seed test data into a user or group namespace.

The Data Seeder uses FactoryBot in the backend which makes maintenance straightforward. When a Model changes, FactoryBot already reflects the change.

Docker Setup

Prerequisites

  • Docker installation

Steps

Run a GitLab container

Run and wait for the container to start. The container has started completely when you see the login page at http://localhost:8080.

With GDK 
$ docker run \
    -d \
    -v ./scripts/data_seeder:/opt/gitlab/embedded/service/gitlab-rails/scripts/data_seeder \
    -v ./ee/db/seeds/data_seeder:/opt/gitlab/embedded/service/gitlab-rails/ee/db/seeds/data_seeder \
    -v ./ee/lib/tasks/gitlab/seed:/opt/gitlab/embedded/service/gitlab-rails/ee/lib/tasks/gitlab/seed \
    --name gitlab \
    gitlab/gitlab-ee:16.7.0-ee.0
Without GDK 
$ docker run \
    --name gitlab \
    -d \
    gitlab/gitlab-ee:16.7.0-ee.0

Get the root password

If you need to fetch the password for the GitLab instance that was spun up, execute the following command and use the password given by the output: 

$ docker exec gitlab cat /etc/gitlab/initial_root_password
5iveL!fe

If you receive cat: /etc/gitlab/initialize_root_password: No such file or directory, please wait for a bit for GitLab to boot and try again.

You can then sign in to http://localhost:8080/users/sign_in using the credentials: root / <Password taken from initial_root_password>

Import the test resources

Because Seeding uses GitLab test resources and given that the GitLab Docker container is meant to be slim, the container does not ship with test resources by default.

By default, the default GitLab branch master is checked out. This means that whatever is “latest” will be checked out. To change this, you can override this ref using the REF environment variable.

Execute the following command to provide test resources (namely, FactoryBot Factories) for the Seeder to use. 

$ docker exec gitlab bash -c "wget -O - https://gitlab.com/gitlab-org/gitlab/-/raw/master/scripts/data_seeder/test_resources.sh | bash"
# OR check out a specific ref
$ docker exec gitlab bash -c "wget -O - https://gitlab.com/gitlab-org/gitlab/-/raw/master/scripts/data_seeder/test_resources.sh | REF=v16.7.0-ee bash"

Seed the data

IMPORTANT: This step should not be executed until the container has started completely and you are able to see the login page at http://localhost:8080. 

$ docker exec -it gitlab bash -c "cd /opt/gitlab/embedded/service/gitlab-rails; wget -O - https://gitlab.com/gitlab-org/gitlab/-/raw/master/scripts/data_seeder/globalize_gems.rb | ruby; bundle install"
Fetching gems...

$ docker exec -it gitlab gitlab-rake "ee:gitlab:seed:data_seeder[beautiful_data.rb]"
Seeding data for Administrator
..........................

GDK Setup 

$ gdk start db
ok: run: services/postgresql: (pid n) 0s, normally down
ok: run: services/redis: (pid n) 74s, normally down
$ bundle install
Bundle complete!
$ bundle exec rake db:migrate
main: migrated
ci: migrated

Run

The ee:gitlab:seed:data_seeder Rake task takes one argument. :file. 

$ bundle exec rake "ee:gitlab:seed:data_seeder[beautiful_data.rb]"
Seeding data for Administrator
....

:file

Where :file is the file path. (This path reflects relative .rb, .yml, or .json files located in ee/db/seeds/data_seeder, or absolute paths to seed files.)

Develop

The Data Seeder uses FactoryBot definitions from spec/factories which …

  1. Saves time on development
  2. Are easy-to-read
  3. Are easy to maintain
  4. Do not rely on an API that may change in the future
  5. Are always up-to-date
  6. Execute on the lowest-level (ActiveRecord) possible to create data as quickly as possible

From the FactoryBot README : factory_bot is a fixtures replacement with a straightforward definition syntax, support for multiple build strategies (saved instances, unsaved instances, attribute hashes, and stubbed objects), and support for multiple factories for the same class, including factory inheritance

Factories reside in spec/factories/* and are fixtures for Rails models found in app/models/*. For example, For a model named app/models/issue.rb, the factory will be named spec/factories/issues.rb. For a model named app/models/project.rb, the factory will be named app/models/projects.rb.

Three parsers currently exist that the GitLab Data Seeder supports. Ruby, YAML, and JSON.

Ruby

All Ruby Seeds must define a DataSeeder class with a #seed instance method. You may structure your Ruby class as you wish. All FactoryBot methods (create, build, create_list) are included in the class automatically and may be called.

The DataSeeder class contains the following instance variables defined upon seeding:

  • @seed_file - The File object.
  • @owner - The owner of the seed data.
  • @name - The name of the seed. This is the seed file name without the extension.
  • @group - The root group that all seeded data is created under.
  • @logger - The logger object to log output. Logging output may be found in log/data_seeder.log. 
# frozen_string_literal: true

class DataSeeder
  def seed
    my_group = create(:group, name: 'My Group', path: 'my-group-path', parent: @group)
    @logger.info "Created #{my_group.name}" #=> Created My Group

    my_project = create(:project, :public, name: 'My Project', namespace: my_group, creator: @owner)
  end
end

YAML

The YAML Parser is a DSL that supports Factory definitions and allows you to seed data using a human-readable format. 

name: My Seeder
groups:
  - _id: my_group
    name: My Group
    path: my-group-path

projects:
  - _id: my_project
    name: My Project
    namespace_id: <%= groups.my_group.id %>
    creator_id: <%= @owner.id %>
    traits:
      - public

JSON

The JSON Parser allows you to house seed files in JSON format. 

{
  "name": "My Seeder",
  "groups": [
    { "_id": "my_group", "name": "My Group", "path": "my-group-path" }
  ],
  "projects": [
    {
      "_id": "my_project",
      "name": "My Project",
      "namespace_id": "<%= groups.my_group.id %>",
      "creator_id": "<%= @owner.id %>",
      "traits": ["public"]
    }
  ]
}

Logging

When running the Data Seeder, the default level of logging is set to “information”.

You can override the logging level by specifying GITLAB_LOG_LEVEL=<level>. 

$ GITLAB_LOG_LEVEL=debug bundle exec rake "ee:gitlab:seed:data_seeder[beautiful_data.rb]"
Seeding data for Administrator
......

$ GITLAB_LOG_LEVEL=warn bundle exec rake "ee:gitlab:seed:data_seeder[beautiful_data.rb]"
Seeding data for Administrator
......

$ GITLAB_LOG_LEVEL=error bundle exec rake "ee:gitlab:seed:data_seeder[beautiful_data.rb]"
......

Taxonomy of a Factory

Factories consist of three main parts - the Name of the factory, the Traits and the Attributes.

Given: create(:iteration, :with_title, :current, title: 'My Iteration')

   
:iteration This is the Name of the factory. The filename will be the plural form of this Name and reside under either spec/factories/iterations.rb or ee/spec/factories/iterations.rb.
:with_title This is a Trait of the factory. See how it’s defined.
:current This is a Trait of the factory. See how it’s defined.
title: ‘My Iteration’ This is an Attribute of the factory that is passed to the Model for creation.

Examples

In these examples, you will see an instance variable @owner. This is the root user (User.first).

Create a Group 

my_group = create(:group, name: 'My Group', path: 'my-group-path')

Create a Project 

# create a Project belonging to a Group
my_project = create(:project, :public, name: 'My Project', namespace: my_group, creator: @owner)

Create an Issue 

# create an Issue belonging to a Project
my_issue = create(:issue, title: 'My Issue', project: my_project, weight: 2)

Create an Iteration 

# create an Iteration under a Group
my_iteration = create(:iteration, :with_title, :current, title: 'My Iteration', group: my_group)

Frequently encountered issues

Username or email has already been taken

If you see either of these errors:

  • ActiveRecord::RecordInvalid: Validation failed: Email has already been taken
  • ActiveRecord::RecordInvalid: Validation failed: Username has already been taken

This is because, by default, our factories are written to backfill any data that is missing. For instance, when a project is created, the project must have somebody that created it. If the owner is not specified, the factory attempts to create it.

How to fix

Check the respective Factory to find out what key is required. Usually :author or :owner. 

# This throws ActiveRecord::RecordInvalid
create(:project, name: 'Throws Error', namespace: create(:group, name: 'Some Group'))

# Specify the user where @owner is a [User] record
create(:project, name: 'No longer throws error', owner: @owner, namespace: create(:group, name: 'Some Group'))
create(:epic, group: create(:group), author: @owner)

parsing id "my id" as "my_id"

See specifying variables

id is invalid

Given that non-Ruby parsers parse IDs as Ruby Objects, the naming conventions of Ruby must be followed when specifying an ID.

Examples of invalid IDs:

  • IDs that start with a number
  • IDs that have special characters (-, !, $, @, `, =, <, >, ;, :)

ActiveRecord::AssociationTypeMismatch: Model expected, got … which is an instance of String

This is a limitation for the seeder.

See the issue for allowing parsing of raw Ruby objects.

YAML Factories

Generator to generate n amount of records

Group Labels 

group_labels:
  # Group Label with Name and a Color
  - name: Group Label 1
    group_id: <%= @group.id %>
    color: "#FF0000"

Group Milestones 

group_milestones:
  # Past Milestone
  - name: Past Milestone
    group_id: <%= @group.id %>
    group:
    start_date: <%= 1.month.ago %>
    due_date: <%= 1.day.ago %>

  # Ongoing Milestone
  - name: Ongoing Milestone
    group_id: <%= @group.id %>
    group:
    start_date: <%= 1.day.ago %>
    due_date: <%= 1.month.from_now %>

  # Future Milestone
  - name: Ongoing Milestone
    group_id: <%= @group.id %>
    group:
    start_date: <%= 1.month.from_now %>
    due_date: <%= 2.months.from_now %>

Quirks

  • You must specify group: and have it be empty. This is because the Milestones factory manipulates the factory in an after(:build). If this is not present, the Milestone cannot be associated properly with the Group.

Epics 

epics:
  # Simple Epic
  - title: Simple Epic
    group_id: <%= @group.id %>
    author_id: <%= @owner.id %>

  # Epic with detailed Markdown description
  - title: Detailed Epic
    group_id: <%= @group.id %>
    author_id: <%= @owner.id %>
    description: |
      # Markdown

      **Description**

  # Epic with dates
  - title: Epic with dates
    group_id: <%= @group.id %>
    author_id: <%= @owner.id %>
    start_date: <%= 1.day.ago %>
    due_date: <%= 1.month.from_now %>

Variables

Each created factory can be assigned an identifier to be used in future seeding.

You can specify an ID for any created factory that you may use later in the seed file.

Specify a variable

You may pass an _id attribute on any factory to refer back to it later in non-Ruby parsers.

Variables are under the factory definitions that they reside in. 

---
group_labels:
  - _id: my_label #=> group_labels.my_label

projects:
  - _id: my_project #=> projects.my_project

Variables:

note
It is not advised, but you may specify variables with spaces. These variables may be referred back to with underscores.

Referencing a variable

Given a YAML seed file: 

---
group_labels:
  - _id: my_group_label #=> group_labels.my_group_label
    name: My Group Label
    color: "#FF0000"
  - _id: my_other_group_label #=> group_labels.my_other_group_label
    color: <%= group_labels.my_group_label.color %>

projects:
  - _id: my_project #=> projects.my_project
    name: My Project

When referring to a variable, the variable refers to the already seeded models. In other words, the model’s id attribute will be populated.