GitOps using Argo CD

When it comes to GitOps, Argo CD is an established approach for consuming Kubernetes objects from a Git repository.

This guide is using the following repository as reference: https://github.com/rigdev/argocd-demo

Configuring Rig

The Rig Platform will by default apply the Kubernetes objects directly through the Kubernetes API. However, for a GitOps setup, this behavior can be changed. The first step is thus to follow the guide here to write the changes to your Git repository instead of Kubernetes directly.

When configuring the cluster, here is an example of using our GitHub repository:

Helm values - Platform

rig:
  clusters:
    my-cluster:
      git:
        url: https://github.com/rigdev/argocd-demo
        branch: main
        path_prefix: "apps/{{ .Cluster.Name }}/{{ .Project.Name }}/{{ .Capsule.Name }}/"

The path_prefix is a template for where Capsule-specific resources will be located - this will be important later when configuring Argo CD.

info

It's recommended to include the cluster-name in the path_prefix path, to allow your repository to be consumed by multiple cluster.

Setup Repo in Argo CD

In Argo CD, using either the Dashboard, CLI or Kubernetes resources directly, configure you repository. It's recommended to configured it with GitHub webhooks, to accelerate the time from making changes until Argo CD picks them up. Without webhooks, the default delay is 3 minutes.

Create Rig Application

Next is to create an Application in Argo CD. The Application should be a "directory" type, that can consume raw Kubernetes manifests from a directory - in this case a directory in our repository.

info

It's also possible to utilize the Kustomize Application type, as described here.

The raw Application should look similar to this:

Argo CD Application

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: rig
spec:
  destination:
    server: https://kubernetes.default.svc
  source:
    path: apps/
    repoURL: https://github.com/rigdev/argocd-demo.git
    targetRevision: HEAD
    directory:
      recurse: true
      include: '*'
  project: default
  syncPolicy:
    automated:
      prune: true

The path we are interested in, should match our path_prefix from above. Anything within this subpath will be included, and in our case that would indeed be anything designated my-cluster cluster.

A few additional notes:

• recurse enables recursive mode, which means all .yaml files in all sub-folders will be consumed.
• prune is required to ensure Argo CD deletes resources, as they are deleted by the Rig Platform.

With that in place, the Application should now be able to sync the repository and apply our Capsule resources.

Deploying from Rig

As seen in the video below, changes to a Capsule are now stored in a git repository, that Argo CD are picking up and applying for you.