Learn Kubernetes From Scratch:
Beginner to CKA (2026)
Architecture explained, core objects decoded, a kubectl cheatsheet you'll actually use, production patterns, and a 12-week study plan to pass the CKA on your first attempt.
Before You Start
Kubernetes is a container orchestrator. If you don't know what containers are, start with Docker first. Specifically, you need to be comfortable with:
- ✓Building Docker images with a Dockerfile
- ✓Running containers and mapping ports
- ✓Pushing images to a registry (Docker Hub, ECR, GCR)
- ✓Basic Linux commands and file system navigation
Kubernetes Architecture in Plain English
A Kubernetes cluster has two types of machines: the control plane (the brain) and worker nodes (the muscle).
The single entry point for all operations. kubectl talks to this.
Distributed key-value store. The only source of truth for cluster state.
Assigns new pods to nodes based on resource availability and constraints.
Watches cluster state and drives it toward desired state (ReplicaSet controller, etc.).
Agent on every node. Receives pod specs from API server and runs containers.
Maintains iptables / eBPF rules for Service networking on each node.
containerd (or CRI-O). Actually runs the container processes.
Network plugin (Calico, Cilium, Flannel) that gives pods their IPs.
The 10 Kubernetes Objects You Must Know
Pod
The smallest deployable unit — one or more containers sharing network and storage.
Deployment
Manages a ReplicaSet. Handles rolling updates, rollbacks, and desired replica count.
Service
Stable network endpoint for pods. Types: ClusterIP, NodePort, LoadBalancer.
Ingress
HTTP/HTTPS routing. Maps hostnames and paths to backend Services.
ConfigMap
Non-sensitive config data mounted into pods as env vars or files.
Secret
Base64-encoded sensitive data (passwords, tokens, TLS certs).
PersistentVolumeClaim
Request for durable storage. K8s binds it to a PersistentVolume.
HorizontalPodAutoscaler
Auto-scales Deployment replicas based on CPU/memory or custom metrics.
NetworkPolicy
Firewall rules for pods. Default-deny, then allow only what's needed.
ServiceAccount
Identity for pods to authenticate to the K8s API and cloud services.
kubectl Commands You'll Use Every Day
kubectl get pods -n <ns>List pods in a namespacekubectl describe pod <name>Full detail — events, conditions, resource limitskubectl logs <pod> -fStream pod logs in real timekubectl exec -it <pod> -- bashOpen a shell inside a running containerkubectl apply -f manifest.yamlCreate or update resources from a filekubectl rollout status deploy/<name>Watch a rolling update in progresskubectl rollout undo deploy/<name>Roll back to the previous revisionkubectl top nodesShow CPU/memory usage per nodekubectl top pods --sort-by=memorySort pods by memory consumptionkubectl get events --sort-by=lastTimestampCluster-wide event log, newest lastkubectl drain <node> --ignore-daemonsetsSafely evict all pods before node maintenancekubectl cordon <node>Mark node unschedulable without evicting pods12-Week CKA Study Plan
The CKA is a practical exam — you'll be given a live cluster and 17 tasks to complete in 2 hours. Studying with videos alone won't cut it. Use this 12-week plan with a real cluster.
Week
1–2
Core concepts
Architecture, etcd, API server, kubectl setup, first Pod + Deployment
Week
3–4
Workloads
ReplicaSets, DaemonSets, StatefulSets, Jobs, CronJobs
Week
5–6
Networking
Services, DNS, Ingress, NetworkPolicy, Calico/Cilium basics
Week
7–8
Storage & Config
PVs, PVCs, StorageClass, ConfigMaps, Secrets, Vault integration
Week
9–10
Security (CKS prep)
RBAC, PSA, image scanning, audit logging, runtime security
Week
11
Cluster admin
kubeadm, upgrades, etcd backup/restore, node troubleshooting
Week
12
Exam simulation
killer.sh practice exams × 3, time management, weak area drills
Frequently Asked Questions
How long does it take to learn Kubernetes?
With Docker basics already in place, expect 8–12 weeks to be comfortable running workloads. CKA exam readiness takes 12–16 weeks of consistent practice on a live cluster.
Do I need Docker before Kubernetes?
Yes. You need to understand container images, building with Dockerfiles, and pushing to a registry. Without this foundation, Kubernetes concepts won't make practical sense.
Is Kubernetes hard to learn?
It has a steep initial curve due to many moving parts. Once you grasp the declarative model — you describe desired state, K8s ensures it — the rest follows logically. Hands-on practice is critical.
What cloud should I use to practice Kubernetes?
For learning, use minikube or kind locally (free). For production-like practice, use EKS on AWS, GKE on GCP, or AKS on Azure — all have free trial credits. GKE Autopilot is the easiest managed K8s for beginners.
Practice on a real cluster
Stop reading. Start kubectl-ing.
Devloud's Kubernetes batches give you a live EKS cluster from day one. You break things. You fix things. You get hired.