Just-in-time Nodes for Any AWS EKS Cluster: Karpenter
summary
Karpenter runs on EKS and replaces Cluster Autoscaler with a groupless provisioning model that talks to the EC2 Fleet API to pick instance types just-in-time. It mixes Spot and On‑Demand, honors pod resource requests and node selectors, and adds consolidation and TTL‑after‑empty logic to shrink efficiently. Demo provisions nodes, schedules pods, and then consolidates and deletes larger nodes—claiming a target to launch thousand pods within 30 seconds—and the live demo shows scaling up and down working end‑to‑end.
AI-written summary · about
chapters (7)
- 0:00Karpenter for EKS auto scaling
- 1:59Kubernetes auto scaling challenges
- 16:14What is Karpenter and how it works
- 22:10Karpenter’s groupless provisioning
- 35:41Key takeaways for Karpenter
- 39:14Live demo: Karpenter install and use
- 59:52Summary and Q&A
related talks
19mJust in time Worker Nodes with Karpenter
31mKubernetes workloads with Karpenter & Amazon EKS
22mUse Karpenter to scale, optimize & upgrade K8s
36mGroupless Autoscaling with Karpenter
22mKarpenter Efficient scaling of Kubernetes clusters
18mKarpenter : Your trusted Companion for Kubernetes Scaling
27mHarness the power of Karpenter to scale, optimize & upgrade K8s
34mReduce Costs and Optimize on Your Cloud with Karpenter