Skip to content

Kubernetes introduction: start here

Kubernetes (K8s) is an open-source container orchestration platform. It automates deploying, scaling, and managing containerized applications across a cluster of machines. This page explains the concepts and architecture, then gives the reading path for the section. The current minor release in November 2026 is Kubernetes 1.37, and the project maintains release branches for the three most recent minor releases.

Docker solves packaging. Kubernetes solves operations at scale.

Without orchestration you face:

  • Manual restarts when containers crash
  • No built-in load balancing across replicas
  • No rolling deployments without downtime
  • No automatic scaling under load

Kubernetes handles all of this declaratively β€” you describe desired state, and the control plane makes it reality.

Kubernetes does not run Docker Engine itself as a runtime: built-in Docker support (dockershim) was removed in Kubernetes 1.24, and nodes use a Container Runtime Interface (CRI) runtime such as containerd or CRI-O. Images built with Docker still run unchanged, because they are standard OCI images.

Control Plane (manages the cluster)
β”œβ”€β”€ API Server β€” single entry point for all operations
β”œβ”€β”€ etcd β€” distributed key-value store for cluster state
β”œβ”€β”€ Scheduler β€” assigns pods to nodes based on resources
└── Controller Manager β€” reconciliation loops (deployments, replicas, etc.)
Worker Nodes (run the workloads)
β”œβ”€β”€ kubelet β€” agent that runs pods on the node
β”œβ”€β”€ kube-proxy β€” network rules and service routing
└── Container runtime β€” containerd or CRI-O (any CRI-compatible runtime)
Concept What it is
Pod Smallest deployable unit β€” one or more containers
Node A machine (VM or physical) in the cluster
Cluster One control plane + multiple worker nodes
Namespace Logical partition within a cluster
Deployment Manages replicas and rolling updates of pods
Service Stable network endpoint in front of pods
ConfigMap / Secret External configuration and sensitive values
Ingress / Gateway HTTP routing from outside the cluster to services (Gateway API is the recommended successor to Ingress)
Concern Docker alone Kubernetes
Container restart on crash Manual or restart policy Automatic
Load balancing Manual with nginx/HAProxy Built-in via Services
Rolling updates Manual scripting kubectl rollout
Auto-scaling Not built-in HorizontalPodAutoscaler
Multi-host networking Complex setup Built-in overlay network

Use plain Docker for local development and single-host deployments. Use Kubernetes when you need multi-node workloads, auto-scaling, or zero-downtime deployments.

Good fit:

  • Microservices with independent scaling requirements
  • High-availability production workloads
  • Teams that ship frequently and need zero-downtime deploys

Might be overkill:

  • Small single-service apps
  • Prototypes or internal tools
  • Teams without operational capacity to maintain a cluster (consider managed alternatives like Cloud Run or Azure Container Apps)

All major clouds offer managed control planes so you only manage worker nodes:

  • GKE β€” Google Kubernetes Engine
  • AKS β€” Azure Kubernetes Service
  • EKS β€” Amazon Elastic Kubernetes Service
Step Page What you get from it
1 This page Concepts, architecture, and when Kubernetes is worth it.
2 Pods and Deployments Pods, probes, resources, rolling updates, rollbacks and autoscaling.
3 Services, Ingress and Gateway Stable networking for Pods and HTTP routing from outside the cluster.
4 Config, namespaces and Helm ConfigMaps, Secrets, namespaces with quotas and RBAC, and Helm charts.
5 kubectl command reference The commands you use every day, grouped by task.