Skip to content

argocd-gitops.md

Latest commit

 

History

History
155 lines (110 loc) · 4.9 KB

argocd-gitops.md

File metadata and controls

155 lines (110 loc) · 4.9 KB

Hello GitOps

Based on original project

Example project to demonstrate GitOps using Kustomize, GitHub Actions and ArgoCD

What we want to achieve

  1. Code change is made to a Python application and pushed to master
  2. GitHub Actions workflow is triggered
  3. Code is first unit tested
  4. If tests are OK, the application is packaged as a Docker image and pushed to DockerHub
  5. The kustomize manifests are edited to reference this new image tag
  6. Kustomize changes are committed and pushed automatically by the workflow
  7. As kustomize files have changed, ArgoCD synchronize the application state with the Kubernetes cluster to deploy or update the app

To implement this scenario, we will use Minikube as the Kubernetes cluster.

Setup Minikube

  • Install Minikube on MacOS with default driver
brew install minikube
minikube start
  • Create the K8s namespaces we will use
kubectl create ns hello-gitops
kubectl create ns argocd
  • Enable the ingress controller (nginx)
minikube addons enable ingress

Setup ArgoCD

Refer to the official setup instructions for a detailed guide. To sum it up

kubectl apply -n argocd -f https://raw.githubusercontent.com/argoproj/argo-cd/stable/manifests/install.yaml

ArgoCD can be interacted with using

  • the web UI
  • the ArgoCD CLI
  • or by editing the CRD yaml

If you want to install the CLI

brew tap argoproj/tap
brew install argoproj/tap/argocd

For this example, we will use the web UI. It can be made accessible using a port forward

kubectl port-forward svc/argocd-server -n argocd 8080:443

Check you have access to the web UI at http://localhost:8080 using your browser

  • Username is admin
  • Default password can be retrieved by running kubectl -n argocd get secret argocd-initial-admin-secret -o jsonpath="{.data.password}" | base64 -d; echo

You are of course encouraged to change this password

Setup DockerHub

Login to your DockerHub account and create a new public repository by clicking Create Repository

Setup GitHub

Fork the repository

  • First, fork this repository by using the GitHub Fork button

  • Then clone your new forked repository

git clone https://github.com/<your github id>/github-actions-samples

Configure GitHub Actions

You will need to expose your DockerHub credentials as GitHub secrets.

  • From your forked project page, click on Settings then Secrets

  • Create two new secrets DOCKER_USERNAME and DOCKER_PASSWORD
  • Check the .github/workflows/ci-argocd-python-app.yml to check parameters and familiarize yourself with the workflow
    • Checkout the code
    • Unit test it
    • Build and push the application as a Docker image
    • Update the Kustomize manifest with the new Docker tag

Take a look at Kustomize

Kustomize manifests are in the kustomize folder. The kustomization.yaml is updated automatically by the GitHub workflow to add a new images transformation setting the right image name and tag.

images:
- name: hello-gitops
  newTag: baadd7d7832f74e3c6d37b3f07b179d7e86c4017
  newName: elisska/hello-gitops

No need to touch these settings as they are updated by the workflow.

Configure ArgoCD

At the end of the workflow, Kustomize manifests are referencing the newly built Docker image. We will configure ArgoCD to observe changes to Kustomize files and update the application in the K8s cluster.

  • Login to ArgoCD
  • See the official documentation for a step by step guide
  • You will need to enter
    • your forked GitHub repository address
    • HEAD or master revision
    • k8s as the Path parameter
    • hello-gitops as namespace
  • Application configuration should look like this

  • If you've pickup up manual synchronization, click the Sync button
  • Your application should be deployed properly in the Minikube cluster

  • You can make some code changes in the project, ArgoCD should catch the changes and redeploy (or detect out of sync if using manual synchronization)

Test the application

Application is reachable on port 8050 inside the cluster. It displays a simple Hello World string with the hostname.

You can reach it through a port forward command

kubectl -n hello-gitops port-forward $(kubectl -n hello-gitops get po -o name | tail -n 1) 8111:8050
curl localhost:8111
Hello, World!