How do you develop a custom plugin for Gitlab CICD? - gitlab

Question

I need to integrate a Gitlab CICD pipeline with a custom Delivery Manager Tool. My pipeline will need to invoke the delivery manager API with some info.
In Jenkins we developed a plugin that provided a pipeline step - void deployEnv (String app, String environment ) - that can be used in the different stages, e.g.:
deployEnv("applicationx", "production")
Is there a way to develop a similar plugin in Gitlab CICD?
Is it possible to invoke a remote URL from a Gitlab CICD pipeline passing some credentials?

Answer 1

The closest analog for this kind of "plugin" in GitLab CI is probably a CI templated job definition. There's maybe a few ways to formulate this. There are a number of methods for abstracting and including job definitions provided by others. Some of the basic tools are: include:, extends:, !reference, "hidden job" keys, and YAML anchors.
Providing reusable templates
If you just need to provide an abstraction for a series of steps, a "hidden key" definition would be the closest to what you want.
Consider the following template YAML. This might be embedded directly in your .gitlab-ci.yml file(s) or you might choose to include it any number of configurations from a remote location using the include: keyword.
In this fictional example, we provide a script step that expects two environment variables to be present: APPLICATION_NAME and DEPLOYMENT_ENV. These variables are used (in this fictional example) to call a remote API passing those values as path parameters. Here, the definitions are provided in a "hidden job" key
.deploy_to_env:
image: curlimages/curl # or otherwise have `curl` in your environment
script:
- |
if [[ -z "$APPLICATION_NAME" || -z "$DEPLOYMENT_ENV" ]]; then
echo "FATAL: you must set APPLICATION_NAME and DEPLOYMENT_ENV variables"
exit 1
fi
- curl -XPOST "https://my-deployment-api.example.com/${APPLICAITON_NAME}/${DEPLOYMENT_ENV}"
Let's assume this yaml file exists in a file named deploy.yml in a project whose path is my-org/templates.
Using templates
Now let's say a pipeline configuration wants to leverage the above definition to deploy an application named applicationx to production.
First, in any case, the project should include: the remote definition (unless you choose to embed it directly -- e.g., copy/paste).
include:
- project: my-org/templates
file: deploy.yml
ref: main # or any git ref, or omit to use default branch
Then you can use the extends: keyword to form a concrete job from the hidden key.
deploy_production:
stage: deploy
extends: .deploy_to_env
variables:
APPLICATION_NAME: "applicationx"
DEPLOYMENT_ENV: "production"
Or, if you want to embed the deployment steps in the middle of other script steps using !reference is useful here.
deploy_production:
stage: deploy
script:
- export APPLICATION_NAME="applicationx"
- export DEPLOY_ENV="production"
# these could also be set in `variables:`
- echo "calling deployment API to deploy ${APPLICATION_NAME} to ${DEPLOY_ENV}"
- !reference [.deploy_to_env, script]
- echo "done"
There's a lot of ways to handle this, these are just two examples.