Skip to content

Manage GitLab configuration as code to make GitLab easily managable, traceable and reproducible.

License

Notifications You must be signed in to change notification settings

Roche/gitlab-configuration-as-code

Repository files navigation

CI/CD Docker Pull count PyPI Documentation Status Last Commit Python versions License

GitLab Configuration as Code (GCasC)

Manage GitLab configuration as code to make it easily manageable, traceable and reproducible.

Table of Contents

Introduction

When configuring your GitLab instance, part of the settings you put in Omnibus or Helm Chart configuration, and the rest you configure through GitLab UI or API. Due to tons of configuration options in UI, making GitLab work as you intend is a complex process.

We intend to let you automate things you do through now UI in a simple way. The Configuration as Code has been designed to configure GitLab based on human-readable declarative configuration files written in Yaml. Writing such a file should be feasible without being a GitLab expert, just translating into code a configuration process one is used to executing in the web UI.

GCasC offers a functionality to configure:

It gives you also a way to:

  • include external files or other Yamls using !include directive
  • inject environment variables into configuration using !env directive into your Yaml configuration.

Visit our documentation site for detailed information on how to use it.

Configuring your GitLab instance is as simple as this:

appearance:
  title: "Your GitLab instance title"
  logo: "http://path-to-your-logo/logo.png"

settings:
  elasticsearch:
    url: http://elasticsearch.mygitlab.com
    username: !env ELASTICSEARCH_USERNAME
    password: !env ELASTICSEARCH_PASSWORD
  recaptcha_enabled: yes
  terms: '# Terms of Service\n\n *GitLab rocks*!!'
  plantuml:
    enabled: true
    url: 'http://plantuml.url'

instance_variables:
  anotherVariable: 'another value'
  MY_VARIABLE:
    value: !env MY_VARIABLE
    protected: false
    masked: true

features:
  - name: sourcegraph
    value: true
    groups:
      - mygroup1
    projects:
      - mygroup2/myproject
    users:
      - myuser

license:
  starts_at: 2019-11-17
  expires_at: 2019-12-17
  plan: premium
  user_limit: 30
  data: !include gitlab.lic

Note: GCasC supports only Python 3+. Because Python 2.7 end of life is January 1st, 2020 we do not consider support for Python 2.

Quick start

Here you will learn how to quickly start with GCasC.

Important! Any execution of GCasC may override properties you define in your Yaml files. Don't try it directly on your production environment.

Visit our documentation site for detailed information on how to use it.

Configure client

You can configure client in two ways:

  • using configuration file:
    [global]
    url = https://gitlab.yourdomain.com
    ssl_verify = true
    timeout = 5
    private_token = <personal_access_token>
    api_version = 4
    
    By default GCasC is trying to find client configuration file in following paths:
    "/etc/python-gitlab.cfg",
    "/etc/gitlab.cfg",
    "~/.python-gitlab.cfg",
    "~/.gitlab.cfg",
    

B You can provide a path to your configuration file in GITLAB_CLIENT_CONFIG_FILE environment variable.

  • using environment variables:
    GITLAB_CLIENT_URL=<gitlab_url> # path to GitLab, default: https://gitlab.com
    GITLAB_CLIENT_API_VERSION=<gitlab_api_version> # GitLab API version, default: 4
    GITLAB_CLIENT_TOKEN=<personal_access_token> # GitLab personal access token
    GITLAB_CLIENT_SSL_VERIFY=<ssl_verify> # Flag if SSL certificate should be verified, default: true

You can combine both methods and configuration settings will be searched in the following order:

  • configuration file
  • environment variables (due to limitations in python-gitlab if using configuration file only GITLAB_CLIENT_TOKEN environment variable will be used)

Personal access token is mandatory in any client configuration approach and you can configure your it by following these instructions

Additionally you can customize HTTP session to enable mutual TLS authentication. To configure this, you should provide two additional environment variables:

GITLAB_CLIENT_CONFIG_FILE=<path_to_client_certificate>
GITLAB_CLIENT_KEY=<path_to_client_key>

Prepare GitLab configuration

GitLab configuration must be defined in Yaml file. You can provide a configuration in a single file, or you can split it into multiple Yaml files and inject them.

For information how to prepare GitLab configuration Yaml file visit our documentation site.

For settings configuration, which defines Application Settings, the structure is flexible. For example

settings:
  elasticsearch:
    url: http://elasticsearch.mygitlab.com
    username: elastic_user
    password: elastic_password

and B

settings:
  elasticsearch_url: http://elasticsearch.mygitlab.com
  elasticsearch_username: elastic_user
  elasticsearch_password: elastic_password

are exactly the same and match elasticsearch_url, elasticsearch_username and elasticsearch_password settings. This means you can flexibly structure your configuration Yaml, where a map child keys are prefixed by parent key (here elasticsearch parent key was a prefix for url, username and password keys). You only need to follow available Application Settings.

You can adjust your Yamls by splitting them into multiple or injecting environment variables into certain values using !include or !env directives respectively. Example is shown below:

settings:
  elasticsearch:
    url: http://elasticsearch.mygitlab.com
    username: !env ELASTICSEARCH_USERNAME
    password: !env ELASTICSEARCH_PASSWORD
  terms: !include tos.md

license: !include license.yml

where:

  • settings.elasticsearch.username and settings.elasticsearch.password are injected from environment variables ELASTICSEARCH_USERNAME and ELASTICSEARCH_PASSWORD respectively

  • settings.terms and license are injected from tos.md plain text file and license.yml Yaml file respectively. In this scenario, your license.yml may look like this:

starts_at: 2019-11-17
expires_at: 2019-12-17
plan: premium
user_limit: 30
data: !include gitlab.lic

Run GCasC

To run GCasC you can leverage CLI or Docker image. Docker image is a preferred way, because it is simple and does not require from you installing any additional libraries. Also, Docker image was designed that it can be easily used in your CI/CD pipelines.

When running locally, you may benefit from running GCasC in TEST mode (default mode is APPLY), where no changes will be applied, but validation will be performed and differences will be logged. Just set GITLAB_MODE environment variable to TEST.

export GITLAB_MODE=TEST

CLI

GCasC library is available in PyPI.

To install CLI run pip install gitlab-configuration-as-code. Then you can simply execute

gcasc

//TODO add more information on CLI usage

Currently, CLI is limited and does not support passing any arguments to it, but behavior can only be configured using environment variables. Support for CLI arguments may appear in future releases.

Docker image

Image is available in Docker Hub.

GCasC Docker image working directory is /workspace. Thus you can quickly launch gcasc with:

docker run -v $(pwd):/workspace hoffmannlaroche/gcasc

It will try to find both GitLab client configuration and GitLab configuration in /workspace directory. You can modify the behavior by passing environment variables:

  • GITLAB_CLIENT_CONFIG_FILE to provide path to GitLab client configuration file
  • GITLAB_CONFIG_FILE to provide a path to GitLab configuration file
docker run -e GITLAB_CLIENT_CONFIG_FILE=/gitlab/client.cfg -e GITLAB_CONFIG_FILE=/gitlab/config.yml
-v $(pwd):/gitlab hoffmannlaroche/gcasc

You can also configure a GitLab client using environment variables. More details about the configuration of GitLab client are available in this documentation.

Examples

We provide a few examples to give you a quick starting place to use GCasC. They can be found in examples directory.

  1. gitlab.cfg is example GitLab client file configuration.
  2. basic is an example GitLab configuration using a single configuration file.
  3. environment_variables shows how environment variables can be injected into GitLab configuration file using !env directive.
  4. modularized shows how you can split single GitLab configuration file into smaller and inject files containing static text using !include directive.

Building

Docker image

Use make to build a basic Docker image quickly.

make docker-build

When using make you can additionally pass DOCKER_IMAGE_NAME to change default gcasc:latest to another image name:

make docker-build DOCKER_IMAGE_NAME=mygcasc:1.0

To get more control over builds you can use docker build directly:

docker builds -t gcasc[:TAG] .

Dockerfile comes with two build arguments you may use to customize your image by providing --build-arg parameter to docker build command:

  • GCASC_PATH defines the path where GCasC library will be copied. Defaults to /opt/gcasc.
  • WORKSPACE defines a working directory when you run GCasC image. Defaults to /workspace.

Python package

Use make to build source distribution (sdist), Wheel binary distribution and Sphinx documentation.

make build

Both source and Wheel distributions will be placed in dist directory. Documentation page will be placed in build/docs directory.

Remember to run tests before building your distribution!

Testing

Before submitting a pull request make sure that the tests still succeed with your change. Unit tests run using Github Actions and passing tests are mandatory to get merge requests accepted.

You need to install tox to run unit tests locally:

# run the unit tests for python 3, python 2, and the flake8 tests:
tox

# run tests in one environment only:
tox -e py37

# run flake8 linter and black code formatter
tox -e flake

# run black code formatter
tox -e black

Instead of using tox directly, it is recommended to use make:

# run tests
make test

# run flake8 linter and black code formatter
make lint

Contribution

Everyone is warm welcome to contribute!

Please make sure to read the Contributing Guide and Code of Conduct before making a pull request.

Contributors

filipowm
Mateusz
randombenj
Benj Fassbind
11mariom
Mariusz Kozakowski

License

Project is released under Apache License, Version 2.0 license.