Ephemeral Clusters as a Service with vcluster, ClusterAPI and ArgoCD

Session @KubeconEU 2023 in Amsterdam: https://sched.co/1HyXe
Session @Open Source Summit 2023 in Vancouver: https://sched.co/1K5IB

Overview

Welcome to Ephemeral Clusters as a Service with vClusters, ClusterAPI and ArgoCD. GitOps has rapidly gained popularity in recent years, with its many benefits over traditional CI/CD tools. However, with increased adoption comes the challenge of managing multiple Kubernetes clusters across different cloud providers. At scale, ensuring observability and security across all clusters can be particularly difficult.

This repository demonstrates how open-source tools, such as ClusterAPI, ArgoCD, and Prometheus+Thanos, can be used to effectively manage and monitor large-scale Kubernetes deployments. We will walk you through a sample that automates the deployment of several clusters and applications securely and with observability in mind.

Prerequisites

For our sample will be using Azure Kubernetes Service (AKS). Before starting, you will need the following:

An Azure account (already logged in with the Azure CLI)
Azure CLI download
Helm CLI and envsubst
Optional but recommented (and the instructions below assume you have one) A working DNS zone in Azure, to use proper DNS names and automatic certifcates provisioning with Let'sEncrypt

Everything else is installed via ArgoCD, so no need for any extra CLI!

Step 1: Create an ArgoCD management cluster with AKS

To create a new management cluster in AKS, run the following commands. Otherwise, if you already have an existing AKS cluster, you can skip this step and proceed to connecting to the existing AKS cluster. Change accoring to your liking, specially the AZURE_DNS_ZONE:

export AZURE_SUBSCRIPTION_ID=$(az account show --query id -o tsv)
az account set --subscription $AZURE_SUBSCRIPTION_ID

export CLUSTER_RG=management
export CLUSTER_NAME=gru
export LOCATION=southcentralus
export IDENTITY_NAME=gitops$RANDOM
export NODE_COUNT=2
export AZ_AKS_VERSION=1.25.6
export AZURE_DNS_ZONE=kube101.dev
export AZURE_DNS_ZONE_RESOURCE_GROUP=dns

Create a resource group for your AKS cluster with the following command, replacing with a name for your resource group and with the Azure region where you want your resources to be located:

az group create --name $CLUSTER_RG --location $LOCATION

To use automatic DNS name updates via external-dns, we need to create a new managed identity and assign the role of DNS Contributor to the resource group containg the zone resource

export IDENTITY=$(az identity create  -n $IDENTITY_NAME -g $CLUSTER_RG --query id -o tsv)
export IDENTITY_CLIENTID=$(az identity show -g $CLUSTER_RG -n $IDENTITY_NAME -o tsv --query clientId)

echo "Sleeping a bit (35 seconds) to let AAD catch up..."
sleep 35

export DNS_ID=$(az network dns zone show --name $AZURE_DNS_ZONE \
  --resource-group $AZURE_DNS_ZONE_RESOURCE_GROUP --query "id" --output tsv)

az role assignment create --role "DNS Zone Contributor" --assignee $IDENTITY_CLIENTID --scope $DNS_ID

Create an AKS cluster with the following command:

az aks create -k $AZ_AKS_VERSION -y -g $CLUSTER_RG \
    -s Standard_B4ms -c $NODE_COUNT \
    --assign-identity $IDENTITY --assign-kubelet-identity $IDENTITY \
    --network-plugin kubenet -n $CLUSTER_NAME

Connect to the AKS cluster:

az aks get-credentials --resource-group $CLUSTER_RG --name $CLUSTER_NAME

Verify that you can connect to the AKS cluster:

kubectl get nodes

Step 2: Install ArgoCD

ArgoCD is an open-source continuous delivery tool designed to simplify the deployment of applications to Kubernetes clusters. It allows developers to manage and deploy applications declaratively, reducing the amount of manual work involved in the process. With ArgoCD, developers can automate application deployment, configuration management, and application rollouts, making it easier to deliver applications reliably to production environments.

You can install ArgoCD on your Kubernetes cluster by running the following commands in your terminal or command prompt. These commands will download and install ArgoCD on your cluster, allowing you to use it for GitOps-based continuous delivery of your applications

NOTE: Make sure to update the values for ingress hostname in the various helm charts under gitops/management folder; Also make sure to update the external-dns values to include your tenantId and subscriptionId. we will update the readme when we find a better way to dynamically inject these values into the helm charts deployed by ArgoCD

Add ArgoCD Helm Repo:

helm repo add argo https://argoproj.github.io/argo-helm
helm repo update

Edit the gitops/management/argocd/argocd-values.yaml with your hostname and domain name for ArgoCD ingress, then install ArgoCD:

helm upgrade -i -n argocd \
  --version 5.29.1 \
  --create-namespace \
  --values gitops/management/argocd/argocd-values.yaml \
  argocd argo/argo-cd

helm upgrade -i -n argocd \
  --version 0.0.9\
  --create-namespace \
  --values argocd-initial-objects.yaml \
  argocd-apps argo/argocd-apps

Verify that ArgoCD is running:

kubectl get pods -n argocd

Access the ArgoCD web UI by running the following command, and then open the URL in a web browser (ingress, external-dns and cert-manager take care of certificates and DNS hostname resolution):

open https://argocd.$AZURE_DNS_ZONE

NOTE: ArgoCD is in read-only mode for anonymous users, that should be enough to monitor the installaation progress, but if you want to change things, retrieve the secret with: kubectl get secret -n argocd argocd-initial-admin-secret -o=jsonpath='{.data.password}'| base64 -D

Step 3: Bootstrap Management Cluster with ClusterAPI

To initialize the AKS cluster with Cluster API and turn it into the management cluster, follow these instructions. Once initialized, the management cluster will allow you to control and maintain a fleet of ephemeral clusters. Unfortunately this part cannot be automated via ArgoCD just yet, although a promising effort is made in the capi-operator repository:

# Run the script, passing the namespace as a parameter (the Azure Managed Identity for the workload clusters)

./capz-init.sh 

# Check the providers
kubectl get providers.clusterctl.cluster.x-k8s.io -A

Step 4: Deploy clusters via Pull Requests

Open a PR against your main branch, modifying the appset-capz.yaml or appset-vcluster.yaml to deploy your ephemeral clusters!