TOPIC CLOUD
Node Autoprovisioning just-in-time Kubernetes node provisioning on AWS
IN 10 SECONDS
Instead of ordering a fixed set of servers in advance, you let the workload declare what it needs. When a pod is waiting for a home, the system instantly rents the exact size server from the cloud provider, boots it, and runs the work.
GOTCHA If you don't set boundaries on what the system can rent, a minor traffic spike could trigger it to provision massive, expensive servers that blow through your budget.
HOW A JUST-IN-TIME NODE BOOTS
01 Your Pod says: 'I need 2 CPUs and 4GB of RAM to run.'
02 Provisioner looks at the cloud's catalog to find the cheapest server that fits.
03 Cloud API receives the order and boots the new virtual machine in seconds.
04 Cluster joins the new server to the pool and places your pod on it.
POKE IT YOURSELF
kubectl get nodeclass,nodepool -A — list Karpenter configuration resources
kubectl logs -f -n karpenter -l app.kubernetes.io/name=karpenter — watch Karpenter scale decisions
Drill this topic →
~70 sec read