Cluster architecture
A centron Kubernetes cluster consists of a Control Plane powered by centron and one or more Node Pools in which your workloads run. This page describes how these components interact and which of them fall under your responsibility.
Overview
┌──────────────────────────────────────────────────────────┐
│ Control Plane (von centron betrieben) │
│ kube-apiserver · etcd · scheduler · controller-manager │
│ optional repliziert (High Availability) │
└───────────────────────────┬──────────────────────────────┘
│ Kubernetes API
┌───────────────────────────┴──────────────────────────────┐
│ Node Pool 1 │ Node Pool 2 │
│ ┌────────┐ ┌────────┐ │ ┌────────┐ │
│ │ Node │ │ Node │ ... │ │ Node │ ... │
│ │ kubelet│ │ kubelet│ │ │ kubelet│ │
│ │ Pods │ │ Pods │ │ │ Pods │ │
│ └────────┘ └────────┘ │ └────────┘ │
└──────────────────────────────────────────────────────────┘
alle Nodes im selben VPC-Netzwerk
Control Plane
The control plane is the brain of the cluster. It receives your requests via the Kubernetes API, stores the cluster’s state and ensures that the actual state matches the desired state.
centron operates, updates and monitors these components on your behalf. You access them exclusively via the API endpoint – SSH access to the control plane is neither required nor provided for.
The Control Plane is included in the price; you are only charged for your worker nodes. By default, it runs as a single instance. For production clusters, you can have it replicated; see HighAvailability.
Managed Components describes in detail which components centron manages.
Node Pools
A node pool is a group of worker nodes with identical configurations – the same instance type and the same size. All nodes in a pool are managed, scaled and updated together.
A cluster can contain several node pools. This makes it possible to run workloads of different sizes within the same cluster without having to configure all nodes to be the same size.
For each node pool, configure the following:
- Instance type: One of the types C0 to C8; see Select plan.
- Number of nodes: A fixed number of nodes.
- Autoscaling: Alternatively, a minimum and maximum number between which the cluster autoscaler adjusts the pool size based on utilisation.
Worker Nodes
Each worker node is a virtual machine on which your pods run. The nodes run the kubelet, the container runtime and the pre-installed centron Metrics Agent, which provides the utilisation data for the Analytics tab.
Nodes pass through the states Provisioning, Running and Deleting. You can selectively restart (Recycle) or remove individual nodes – the pool provides replacements as required.
Network layers
A cluster uses three separate address ranges:
| Network | Purpose |
|---|---|
| Node network | Addresses of the worker nodes in the VPC. The nodes use these addresses to communicate with one another and with other resources in the same data centre. |
| Pod network | Address range from which individual pods obtain their IP addresses. |
| Service network | Address range for ClusterIP services, via which pods can be reliably addressed. |
All resources in a data centre belong to the same VPC network and can be accessed via private IP addresses. You can view the current address ranges for your cluster in the Overview tab, under the Networking section.
The size of the pod and service networks cannot be changed once the cluster has been created. Plan these areas generously if you anticipate significant growth. When creating the cluster, you can leave the sizing to centron or specify the subnet sizes yourself.
Division of responsibilities
| Area | Responsibility |
|---|---|
| Control plane, including etcd | centron |
| Operating system and Kubernetes components on the nodes | centron (via automatic upgrades) |
| Pool and node provisioning | Customer |
| Workloads, manifests, images | Customer |
| Network policies and RBAC rules within the cluster | Customer |
| Application data backup | Customer |
For a more detailed definition, please see Scope of Support.
Related topics
- High Availability
- Managed Components
- Glossary – key Kubernetes terms explained
- Create a cluster