Skip to main content

Kube-Hetzner Architecture Diagram

Explanation of Components:​

  • User: Interacts with the system, primarily through Terraform.
  • Terraform: Infrastructure as Code tool responsible for provisioning and managing all Hetzner Cloud resources.
  • Hetzner Cloud: The underlying cloud provider.
    • Hetzner Cloud Load Balancer (HC_LB): Used to expose the Kubernetes API server and potentially other services to external traffic.
    • Hetzner Private Network (HC_PN): Provides isolated and secure communication between all cluster nodes.
    • Control Plane Nodes (HC_CP_VM1, HC_CP_VM2, HC_CP_VM3): Virtual Machines hosting the Kubernetes control plane components. An odd number is used for High Availability.
    • Worker Nodes (HC_WK_VM1, HC_WK_VM2): Virtual Machines running user workloads.
    • Hetzner Cloud Volumes (HC_VOL): Block storage volumes used for persistent data.
  • Kubernetes Cluster (k3s): The lightweight Kubernetes distribution.
    • Kubernetes API Server (K8S_API): The front-end for the Kubernetes control plane, exposing the Kubernetes API.
    • etcd (K8S_ETCD): Distributed key-value store used as Kubernetes' backing store for all cluster data.
    • Scheduler (K8S_SCHED): Watches for newly created Pods that have no assigned node, and selects a node for them to run on.
    • Controller Manager (K8S_CTRL): Runs controller processes that regulate the state of the cluster.
    • Kubelet (K8S_KUBELET_CP, K8S_KUBELET_WK): An agent that runs on each node in the cluster, ensuring that containers are running in a Pod.
    • Kube-Proxy (K8S_PROXY): Network proxy that runs on each node, implementing part of the Kubernetes Service concept.
    • CNI (K8S_CNI): Container Network Interface, responsible for pod networking (e.g., Flannel, Calico, Cilium).
    • Storage (K8S_STORAGE): Handles persistent storage for the cluster (e.g., Longhorn, Hetzner CSI Driver).
    • Klipper LoadBalancer (K8S_KLB): k3s's embedded on-metal load balancer for services of type LoadBalancer.

Flow of Operations:​

  1. User Interaction: The user defines the desired infrastructure using Terraform configuration files.
  2. Terraform Provisioning: Terraform interacts with the Hetzner Cloud API to provision all necessary resources: Load Balancers, Private Networks, Control Plane VMs, Worker VMs, and Cloud Volumes.
  3. Network Connectivity: All Control Plane and Worker Nodes are connected to the Hetzner Private Network for secure internal communication.
  4. Kubernetes API Access: The Hetzner Cloud Load Balancer exposes the Kubernetes API Server, providing a stable endpoint for kubectl access.
  5. Control Plane Components: Control Plane Nodes host the Kubernetes API Server, etcd, Scheduler, and Controller Manager, forming the brain of the cluster.
  6. Node Agents: Kubelet and Kube-Proxy run on both Control Plane and Worker Nodes to manage pods and network traffic.
  7. Pod Networking: The chosen CNI (Flannel, Calico, or Cilium) enables communication between pods across the cluster.
  8. Persistent Storage: Hetzner Cloud Volumes are integrated via the CSI driver or managed by Longhorn to provide persistent storage for stateful applications.
  9. Internal Load Balancing: Klipper LoadBalancer handles internal service exposure within the cluster.