Stop 09 · Containers & OrchestrationTopics · 4/9

Managed Kubernetes (EKS, GKE, AKS)

In short

Running the Kubernetes control plane yourself means managing etcd backups, certificates and upgrades. Managed Kubernetes hands that hard part to the cloud provider, so you focus on your apps.

Cloud-managed Kubernetes services that handle the complex control plane for you — you still manage worker nodes and workloads, but the provider handles the hardest operational parts of running Kubernetes itself.

The blob character in cricket whites walks out with a bat as a groundsman rolls the pitch.

Think of it like this

Renting the stadium

Your team brings players, kit and a game plan. The stadium's ground staff roll the pitch, fix the floodlights and mow the outfield. You'd never build a stadium to play one league. Managed Kubernetes is the stadium with staff included.

Watch it happen

Who runs which part

Your cluster

Control plane

API server

etcd database

Scheduler

Node group

Worker node

Your pods

Add-ons

1/4A Kubernetes cluster has two halves: the brain that makes decisions and the machines that run containers.

The big ideas

What you need to know

The big three

AWS EKS, Google GKE and Azure AKS. Same Kubernetes API, different consoles, defaults and integrations.

Hands-off modes

GKE Autopilot and EKS Auto Mode also manage the worker nodes, so you pay closer to per pod.

Upgrades still matter

Kubernetes versions go out of support regularly. You still plan and test cluster upgrades.

Costs add up

You pay for nodes, load balancers, storage and often a per-cluster fee. Delete test clusters the same day.

See it in code

Walk through the code

bash

1# GKE Autopilot: Google manages the nodes too2gcloud container clusters create-auto campus --region asia-south13gcloud container clusters get-credentials campus --region asia-south14 5# EKS with eksctl and one managed node group6eksctl create cluster --name campus --region ap-south-1 \7  --nodegroup-name workers --node-type t3.medium --nodes 28 9# Same kubectl either way10kubectl get nodes

1/3One command creates an Autopilot cluster; the second writes your kubeconfig.

asia-south1 and ap-south-1 are both Mumbai. These clusters cost real money per hour, so delete them right after practising.

Save yourself the pain

Common mistakes

Oops

Leaving a practice EKS or GKE cluster running over the weekend.

Do this instead

Set a billing alert first, and delete clusters with eksctl delete cluster or gcloud the same day.

Oops

Assuming managed means you never touch Kubernetes versions.

Do this instead

Track your version's end-of-support date and schedule upgrades, testing in a non-prod cluster first.

Your mission

Same YAML, real cloud

~45 min

Quick check

In a managed Kubernetes service, what does the provider take care of?

Go deeper

1Launch a managed Kubernetes cluster on your primary cloud provider.
2Deploy a simple application to your managed cluster using kubectl.
3Learn what the provider manages (control plane) versus what remains your responsibility (worker nodes, workloads).