Skip to main content

Real use cases + complete FAQ

Summary: Kubernetes is everywhere — from Netflix streaming to 250 million households, to Spotify personalizing your playlists, to OpenAI running ChatGPT. This final lesson presents 6 documented real use cases, answers the 14 most frequent questions about Kubernetes, and guides you towards your next learning step.

1. Use case 1 — Netflix, king of streaming​

Netflix is one of the biggest showcases of Kubernetes in the world.

Lesson: Netflix shows that Kubernetes excels for microservice backends, but does not try to replace specialized CDNs.

2. Use case 2 — Spotify, the development accelerator​

Spotify has been using Kubernetes since 2019 to accelerate its teams' velocity.

Lesson: Kubernetes is not just a technical tool — it is an organizational velocity accelerator when well encapsulated.

3. Use case 3 — OpenAI, ChatGPT at planetary scale​

OpenAI runs ChatGPT and its APIs on gigantic Kubernetes clusters with GPUs.

Lesson: Kubernetes is the de facto standard for large-scale AI too, with extensions to manage special resources (GPU, TPU, InfiniBand).

4. Use case 4 — ING Bank, a regulated migration​

ING (Dutch bank) migrated massively to Kubernetes in a regulated banking environment.

Lesson: Kubernetes is compatible with the most regulated environments (banking, healthcare, defence), provided you choose the right distribution (OpenShift, Rancher) and put the adequate controls in place.

5. Use case 5 — SNCF, chatbots and internal apps​

The SNCF (French national railway) uses Kubernetes for gradual IT modernization.

Lesson: Kubernetes is also used by large public and industrial European players — not only by the American tech giants.

6. Use case 6 — Edge Kubernetes at Chick-fil-A​

Chick-fil-A (American restaurant chain) uses k3s in 2600+ restaurants.

Lesson: Kubernetes goes all the way to the edge with ultra lightweight distributions like k3s — even in an unexpected context like fast food.

7. Application domains where Kubernetes shines​

Emerging domain in 2026: Kubernetes for AI — Kubeflow, Ray Operator, distributed JAX — is becoming a major strategic axis.

8. When Kubernetes is NOT the right choice​

Kubernetes is not always the right solution. Here is when to abstain.

Golden rule: if your application runs well on Docker Compose with fewer than 20 containers, you probably do not need Kubernetes.

9. FAQ — The 14 most frequent questions​

1. Is Kubernetes free?

The Kubernetes project itself (open source under the Apache 2.0 licence) is 100% free. But operating a cluster costs money: hardware/VMs, network, storage, and possibly the managed control plane (EKS 73,GKEStandard73, GKE Standard 72/month, AKS free). Count on 200−500/month∗∗forasmallmanagedproductioncluster,∗∗200-500/month** for a small managed production cluster, **1000-3000/month for a medium cluster, and tens of thousands for a large company.

2. Should I learn Docker before Kubernetes?

Strongly recommended. Kubernetes orchestrates containers — without understanding what a container, an image, a Dockerfile, a volume are, you will be lost. The 40-minute Docker discovery course is enough to approach Kubernetes with confidence.

3. How long to master Kubernetes?

Realistically, without lying:

  • 1-2 weeks to understand the concepts and know how to write basic manifests.
  • 2-3 months of daily practice to be operational in production on standard cases.
  • 1-2 years of intensive practice to master the complete ecosystem (Helm, Argo CD, security, monitoring, networking).
  • A whole career to cover all the exotic cases.

The CKAD certification validates a good level after 2-3 months of learning.

4. Do I need a local cluster to learn?

Absolutely yes. Three free tools:

  • minikube — a complete K8s cluster on your laptop, the most popular.
  • kind — Kubernetes-in-Docker, ultra fast to start.
  • k3d — k3s in Docker, very lightweight.

To test real cloud behaviour, GKE Autopilot is free with the GCP welcome credits ($300).

5. Is YAML mandatory?

Yes for standard manifests. Possible alternatives:

  • JSON — accepted but rarely used (unreadable for humans).
  • Helm charts — YAML with Go templates.
  • Kustomize — YAML with per-environment overlays.
  • Pulumi, CDK8s — write in TypeScript / Python (compiles to final YAML).

In 2026, 80%+ of teams use Helm + Kustomize on top of raw YAML.

6. Does Kubernetes replace Docker?

No, they are complementary:

  • Docker = image builds + local/dev execution.
  • Kubernetes = orchestrates containers at scale.

Since 2020, Kubernetes uses containerd internally to run containers (Docker Engine is no longer used on the runtime side). But Docker remains ubiquitous for image building and local dev.

7. What is the difference between Deployment and StatefulSet?
  • Deployment = for stateless applications. Interchangeable Pods. Cases: APIs, frontends, workers.
  • StatefulSet = for stateful applications. Pods with a stable identity, persistent storage. Cases: databases, message queues, distributed systems like Kafka, Elasticsearch, MongoDB.

Simple rule: if you lose a Pod, is it serious? No → Deployment. Yes → StatefulSet.

8. How to manage sensitive secrets?

Three approaches:

  1. Native Kubernetes Secrets — base64-encoded, encrypted at rest in etcd. Simple but visible to anyone with kubectl get.
  2. External Secrets Operator — Kubernetes dynamically reads from Vault, AWS Secrets Manager, GCP Secret Manager. The 2026 standard.
  3. Sealed Secrets or SOPS — encrypting secrets BEFORE committing them to Git.

Never commit a plain-text secret to Git, even in a private repo.

9. Can K8s run Windows?

Yes, since Kubernetes 1.14 (2019). You can have a mixed cluster with Linux nodes and Windows nodes. Windows Pods run Windows containers (based on Windows Server Core or Nano Server). Azure's AKS offers the most polished Windows experience.

But 90%+ of Kubernetes workloads remain on Linux, notably for performance and cost reasons.

10. Kubernetes vs Docker Compose?

Compose = local and small, Kubernetes = production and large.

  • Docker Compose = 5-20 containers, one machine, dev/staging, simple YAML.
  • Kubernetes = 20-∞ containers, N machines, production, complex YAML + ecosystem.

Good workflow: Docker Compose for local dev, Kubernetes for staging + production. Some tools (compose-cli with ACI, kompose) translate Compose into Kubernetes automatically.

11. How many nodes minimum for a production cluster?

Minimal recommendation:

  • Control plane: 3 nodes (masters) spread over 3 availability zones.
  • Worker nodes: minimum 3 as well, to tolerate the loss of a node without degradation.

On a managed cluster (EKS, GKE, AKS), the control plane is managed by the cloud — you only count the workers. 3 workers minimum for high availability.

12. Is Kubernetes secure by default?

No — you must harden it. Essential points to put in place:

  • RBAC enabled and strict (on by default, but often badly configured).
  • Network Policies to filter traffic between Pods.
  • Pod Security Standards (restricted by default).
  • Secrets encrypted at rest in etcd.
  • Image scanning (Trivy, Grype).
  • cert-manager for HTTPS everywhere.
  • Audit logs enabled.

The Premium Kubernetes course dedicates 3 modules to security, in preparation for the CKS certification.

13. How to migrate an existing application to Kubernetes?

Recommended 5-step approach:

  1. Containerize — write a Dockerfile for your app.
  2. Local Docker Compose — verify it runs in a container.
  3. Write the Kubernetes manifests — Deployment, Service, ConfigMap.
  4. Test on minikube — validate the local K8s behaviour.
  5. Deploy on a dev cluster, then staging, then prod (with GitOps).

Realistic time: 2-4 weeks for a simple application, several months for a complex monolith.

14. Which Kubernetes certification to start with?

Recommended order:

  1. CKAD — Certified Kubernetes Application Developer. The most accessible, dev-oriented. Ideal for validating your basics after this course.
  2. CKA — Certified Kubernetes Administrator. Harder, ops- and cluster-oriented. For DevOps and SRE.
  3. CKS — Certified Kubernetes Security Specialist. The hardest, security. Prerequisite: having the CKA.

Price: 395 USD per certification, valid 3 years. Exam: 2 hours online, hands-on format in a real terminal, with online docs allowed. Pass rate: about 65-70% on the first attempt.

The Premium Kubernetes Course includes targeted preparation for the CKAD.

10. What you learned in this course​

You have reached the end of the 7 Kubernetes discovery lessons. Let's recap what you now know.

Congratulations! You now have the complete Kubernetes culture — the one expected for:

  • A DevOps, SRE, cloud engineer, platform engineer interview.
  • A conversation with your technical team without getting lost.
  • An informed choice between the managed cloud offerings.
  • A POC project with Kubernetes in local dev.

11. Your next step​

Three paths to continue your learning.

Path 1 · Complete your DevOps culture​

Path 2 · Move to intensive practice​

Premium Kubernetes Course — 40+ hours of practice with:

  • Writing real manifests and applying them.
  • Using kubectl in depth (60+ commands).
  • Deploying PostgreSQL, Redis, Kafka with Helm.
  • Advanced GitOps with Argo CD.
  • Security (RBAC, Network Policies, Pod Security Standards).
  • Complete monitoring (Prometheus, Grafana, Loki).
  • Official CKAD preparation.

Path 3 · Get certified​

  • CKAD — developer-oriented, the most accessible to start.
  • CKA — ops-oriented, if you are aiming for SRE or platform engineer.
  • CKS — security, after having the CKA.

Official certification site: training.linuxfoundation.org.

12. One last piece of advice​

Do not stay in theory. Kubernetes is only truly understood once you have typed it on the keyboard. Install minikube or kind tonight. Write your first deployment.yaml within 24 hours. Run kubectl apply. Watch the Pods start.

It is this first direct experience that will transform your abstract understanding into an operational skill. Safe travels!

Remember in 30 seconds​

  • Kubernetes is everywhere: Netflix, Spotify, OpenAI, ING, SNCF, Chick-fil-A.
  • Main use cases: microservices, web with variable traffic, batch, ML/AI, internal platforms, edge.
  • Do NOT use Kubernetes if you have: a small team, a single monolith, constant and low traffic, or a very tight cloud budget.
  • Next step: install minikube and type your first kubectl apply.
  • Recommended certification to start: CKAD.
  • The Premium Kubernetes Course transforms this culture into an operational skill.

Thank you for taking this Kubernetes discovery course! 🎉

Want to continue? Back to the table of contents or move on to the Premium Kubernetes Course.

Last step: take the end-of-course quiz — five corrected and explained questions, three minutes, to verify that the essentials are acquired.