IT 3300 : Virtualization

Cloud & Managed Kubernetes

You've built it the hard way

  • You installed, clustered, networked, and secured k8s by hand
  • That's the best way to understand it
  • In industry, most teams use managed Kubernetes

Managed Kubernetes

  • GKE — Google Kubernetes Engine
  • EKS — Amazon Elastic Kubernetes Service
  • AKS — Azure Kubernetes Service
  • The cloud runs the control plane; you run workloads

What "managed" buys you

  • No control-plane upgrades/patching to babysit
  • Integrated load balancers, storage classes, and IAM
  • Node autoscaling — add/remove worker nodes on demand
  • Backups, logging, monitoring hooks built in

The trade-offs

  • Cost — you pay for the convenience and the managed plane
  • Lock-in — deep cloud integrations are hard to move
  • Less control — you don't own the control plane
  • Standard k8s skills still transfer — that's the point

Creating a cluster (shape of it)

    # GKE
    gcloud container clusters create demo \
      --num-nodes=3 --zone=us-central1-a

    gcloud container clusters get-credentials demo
    kubectl get nodes
  • Same kubectl, same manifests — different control plane

Your manifests just work

  • The Deployments, Services, and Ingress you wrote are portable
  • A cloud LoadBalancer service now really gets an external IP
  • StorageClasses map to real cloud disks
  • Write once, run on any conformant cluster

Watch the bill

  • Cloud is rented — idle clusters still cost money
  • Tear down lab clusters when done
  • Understand free-tier limits before you spin things up

Lab goals

  • Create a small managed cluster (free tier / credits)
  • Deploy your capstone app to it unchanged
  • Delete the cluster and confirm billing stops