Kubernetes Architecture
Start here because every later Kubernetes object is a negotiation with this architecture: desired state goes into the API server, controllers reconcile it, and nodes turn it into running containers.
Topics
- Kubernetes Components - A Kubernetes cluster is not a single daemon. It is a small distributed operating system with an API, a database, reconcilers, and agents on each machine.
- Control Plane - The control plane is the brain and ledger of the cluster. It accepts desired state, stores it, and runs loops that push reality toward that state.
- Nodes And Kubelet - A node is useful only if kubelet can keep reporting health and can ask the container runtime to run Pods.
- Controllers And Reconciliation - Kubernetes works because controllers keep comparing desired state with observed state. That loop is the product.
- Addons And Architecture Choices - The core cluster is intentionally incomplete. DNS, ingress, metrics, storage, and policy are choices you compose.
Editorial Notes
- Keep the official docs as the citation layer.
- Keep this lab opinionated around homelab operations: verification, failure modes, and practical boundaries.
- Avoid copying upstream prose; cite it, then explain the operating model in our own language.
Sources
- Kubernetes Components - source path:
content/en/docs/concepts/overview/components.md, commit 8cc9e19b8eec8d5cf49eacd66f86a81648edb1a0. - Cluster Architecture - source path:
content/en/docs/concepts/architecture/_index.md, commit 8cc9e19b8eec8d5cf49eacd66f86a81648edb1a0. - Nodes - source path:
content/en/docs/concepts/architecture/nodes.md, commit 8cc9e19b8eec8d5cf49eacd66f86a81648edb1a0. - Kubernetes documentation is licensed under CC BY 4.0; these notes are original commentary and link back to the official source.