Publish to GitLab Pages
GitLab is a DevOps platform that provides everything you need to publish your Antora-based documentation to the web. On this page, we’ll explore how to use GitLab to publish your first documentation site using Antora.
GitLab overview
Each GitLab project provides code hosting (a git repository), continuous integration (GitLab CI), and static web hosting with redirect support (GitLab Pages). GitLab is thus suitable to manage your entire Antora-based documentation site end-to-end, from source to published site.
To use GitLab CI/CD for publishing an Antora site, you need:
-
A playbook project, which starts as a local git repository that stores the Antora playbook file.
-
Zero or more content projects, each with a git repository that hosts the content files.
-
A file named .gitlab-ci.yml in the root of the git repository in your playbook project, which provides the CI/CD configuration.
Before proceeding, we recommend studying core CI/CD concepts and the overview of the .gitlab-ci.yml file in the GitLab documentation.
Get started
Begin by creating a new project on GitLab to host your playbook project. Follow the instructions on that page to push your local git repository to GitLab. If your playbook refers to other content source repositories, make sure to push those to GitLab as well.
Next, you’ll need to configure CI/CD. You only need to set up GitLab CI/CD in the default branch of your playbook repository. Antora will fetch content from other repositories that are declared in your playbook automatically.
We’ll start with a basic GitLab CI/CD configuration file that uses the Antora Docker image to build a site with Antora and publish it to GitLab Pages.
image:
name: antora/antora
pages:
stage: deploy
interruptible: true
script:
- antora --fetch --redirect-facility=gitlab --to-dir=public antora-playbook.yml
artifacts:
paths:
- public
Commit the .gitlab-ci.yml file to git and push it to the remote repository. That will trigger the first CI/CD pipeline. If the pipeline succeeds, your site will be accessible at the URL listed on the Pages settings page.
By default, GitLab Pages automatically enables the Use unique domain option, which adds a unique number to the URL where the site is hosted. Chances are, this is not what you want. It’s not only an obscure URL, it can disrupt the use of a custom domain. Follow these steps to correct it:
-
Navigate to the
menu option from the project repository. -
De-select the Use unique domain option.
-
Click Save changes.
The URL for the GitLab Pages site will now follow this pattern:
https://<group-name>.gitlab.io/<project-name>
The site will be public even if the project is private.
You can refer to the playbook project for the Antora demo to see another example of building and publishing an Antora site using GitLab Pages that taps into a few additional capabilities of GitLab CI/CD.
Customize the build
So far, we’ve relied on the Antora Docker image to run Antora in CI/CD. The Antora Docker image only provides the Antora core components. It does not include any extensions.
While using the Antora Docker image is a convenient way to get up and running quickly, we strongly recommend declaring the dependencies of your site within your playbook project (or extending the Docker image). By doing so, you keep your build self-contained and portable. This is especially important if you rely on additional packages such as the @antora/lunr-extension and asciidoctor-kroki.
Let’s assume you have the following package.json file in your playbook project that declares a dependency on Antora, the Antora Lunr Extension, and Asciidoctor Kroki.
{
"name": "my-docs-site",
"description": "My Docs Site",
"private": true,
"devDependencies": {
"antora": "~3.0",
"@antora/lunr-extension": "1.0.0-alpha.8",
"asciidoctor-kroki": "0.15.4"
}
}
You’ll first want to run npm i
to generate the package-lock.json file and commit both files.
With that configuration in place, you can modify your GitLab CI/CD configuration to be based on node:16-alpine, the base image used by Antora Docker. You’ll then need to fetch the dependencies on each build since the base image does not provide Antora.
image:
name: node:16-alpine (1)
variables:
ANTORA_CACHE_DIR: .cache/antora (2)
NODE_OPTIONS: --max-old-space-size=4096 (3)
before_script:
- npm ci (4)
pages:
stage: deploy
interruptible: true
rules:
- if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH (5)
cache:
paths:
- .cache (2)
script:
- npx antora --fetch --redirect-facility=gitlab antora-playbook.yml (6)
artifacts:
paths:
- public (7)
1 | The base image used by Antora Docker, which you can use if you’re installing packages locally to the playbook project. |
2 | Stores the repository and UI bundle cache between runs. |
3 | (optional) Increases the memory reserved for the Node.js to allow Antora to process heavier builds. |
4 | Installs dependencies defined in the package-lock.js file.
Using npm ci instead of npm i ensures that the versions of your dependencies remain stable between runs. |
5 | Only run on the default branch for this repository. |
6 | Call Antora using npx , which will locate and run the antora command installed within the project.
The --fetch flag ensure Antora fetches updates into the cache saved from a previous run. |
7 | The public directory is the predefined folder for publishing as site to GitLab Pages. |
If any of your content repositories are private, you can define a GIT_CREDENTIALS
CI/CD variable that holds the credentials to give Antora access to those repositories in this environment.
You can set up deploy tokens in your content repository to give the CI/CD pipeline in your playbook project (and thus Antora) read-only access to those repositories.
If you want Antora to fail the CI/CD pipeline if there are any warnings or non-fatal errors, add --log-failure-level=warn
to the antora
command.
Alternately, you can set the runtime.log.failure_level
key in the playbook to make it a permanent setting.