Real world example of Kops deployment with 2 environments: lab and test. Use FluxCD to automatically deploy both infrastructure and applications.
The kubernetes cluster deploys on the three AZ of eu-west Region in a single VPC. All instance groups are Spot Fleets that created nodes inside the private subnet. All the nodes are running AL2 images and run on Graviton2 ARM instances.
A master-plane public NLB allow access to the Kubernetes API. The front public NLB will be created using a Kubernetes Ingress resource.
Cilium is the CNI provider deployed in the cluster.
Kops components used:
| Name | Status | Description |
|---|---|---|
| AWS IAM Authentication | ✅ | |
| AWS Cloud Controller Manager | ✅ | |
| AWS Load Balancer Controller | ❌ | A NLB will be created when deploying an Ingress |
| Cluster autoscaler | ❌ | Karpenter is used as a replacement |
| Cert Manager | ❌ | Deployed externally using FluxCD |
| Cilium | ✅ | |
| CoreDNS | ✅ | |
| DNS Controller | ✅ | |
| EBS CSI | ✅ | |
| External VPC | ✅ | |
| FluxCD | ✅ | Custom addon |
| Hubble | ✅ | Deploy Server & Relay |
| Karpenter | ✅ | 🚧 WIP |
| Metric Server | ❌ | Deployed using FluxCD & Prometheus Kube Stack |
| Node local DNS | ✅ | |
| Node termination handler | ✅ | |
| Node Problem Detector | ✅ | |
| Snapshot controller | ❌ |
🚧 Terraform to create Kops S3 state bucket
To start the cluster
make create
make update
To delete the cluster
make delete