Kubernetes doesn’t let us run containers by themselves. They must be wrapped in a pod first.

• Service mesh
• Sidecar
• Don’t allow Scaling.

Architecture

Deployments wrap pods

• One deployment manage one type of pod
  • Can’t have an app and DB in same deployment
  • Would have to make two deployments

Replica-set is an intermediary that actually handles mechanics of the scaling and replication of pods within a deployment

Strategy and Updating a Cluster

replicas: 10000
strategy:
	rollingUpdate:
		maxSurge: 1 # How many extra "instances" allocated during provisioning
		maxUnavailable: 0 # Number you can go below desired state (replicas)
	type: RollingUpdate

Zero downtime operation, if clients can handle it.

Replica Set helps you through it 🙂. We can manipulate them if we really want to. Doesn’t mean we should.

Labels

Labels tell the Deployment which Pods to select.

• The Deployment will manage all pods with a particular label, whether the Replica Set creates them or whether you manually add them to the cluster