Kubernetes (k8s) is a container orchestration system that automates deployment, scaling, and management of containerized applications.

Architecture Overview

  • Control Plane: manages cluster state and scheduling.

    • kube-apiserver: API front end (single source of truth).
    • etcd: key-value store for cluster state.
    • kube-scheduler: assigns pods to nodes.
    • kube-controller-manager: runs controllers (replication, endpoints, nodes).
    • cloud-controller-manager (optional): cloud-provider integration.

  • Nodes (Worker Plane): run workloads.

    • kubelet: agent that manages pods on a node.
    • kube-proxy: implements Service networking (iptables/ipvs).
    • Container runtime: containerd, docker, or CRI-compatible runtime.

  • Add-ons / Services

    • DNS (CoreDNS), ingress controllers, metrics-server, network plugins (CNI).

Key Concepts

  • Pod: smallest deployable unit (one or more containers sharing network and storage).
  • Deployment: declarative updates for Pods and ReplicaSets.
  • Service: stable network endpoint exposing Pods.
  • ConfigMap / Secret: configuration and sensitive data.
  • PersistentVolume (PV) / PersistentVolumeClaim (PVC): storage abstraction.
  • Namespace: virtual cluster partitioning.

Basic Usage (cheat sheet)

# see cluster state
kubectl cluster-info
kubectl get nodes
kubectl get pods -A

# deploy an app
kubectl create deployment myapp --image=nginx:stable
kubectl expose deployment myapp --port=80 --type=ClusterIP

# scale
kubectl scale deployment/myapp --replicas=3

# update image
kubectl set image deployment/myapp myapp=nginx:1.24

# inspect and debug
kubectl describe pod <pod>
kubectl logs deployment/myapp
kubectl exec -it <pod> -- /bin/sh

# persistent storage example
kubectl apply -f - <<EOF
apiVersion: v1
kind: PersistentVolume
metadata:
	name: example-pv
spec:
	capacity:
		storage: 1Gi
	accessModes:
		- ReadWriteOnce
	hostPath:
		path: /tmp/data
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
	name: example-pvc
spec:
	accessModes:
		- ReadWriteOnce
	resources:
		requests:
			storage: 1Gi
EOF

Recommendations

  • Use declarative manifests (kubectl apply -f) and store them in git.
  • Prefer readinessProbes and livenessProbes for production workloads.
  • Use namespaces to separate environments (dev/stage/prod).
  • Monitor cluster health (metrics-server, Prometheus) and logs.

If you want, I can add diagrams, sample manifests for a full app (Service+Deployment+PVC), or a short troubleshooting checklist.