Karpenter Efficient scaling of Kubernetes clusters
summary
Karpenter is a groupless autoscaler that communicates directly with the EC2 API to pick exact instance types and AZs, calculate bin-packing including DaemonSet overhead, and run consolidation simulations that replace or delete nodes. Demo provisions 3,000 pods (250m CPU / 256Mi each) by requesting ~22 nodes and achieving about 93% utilization, shows a 50/50 spot and on-demand split for a 600-pod test, provisions new nodes in roughly 60s, migrates pods during consolidation, and reduced node count automatically while costing about $26/hr in the demo.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:34How Karpenter works
- 4:43Demo overview
- 5:46High-load demo start
- 7:12First environment setup
- 8:16Deploying first application
- 9:24Karpenter logs and scaling
- 10:50Consolidation in action
- 11:56Spot and On-Demand mix
- 13:43Topology spread constraints
- 15:09Scaling 600 pods
- 16:373000 pods efficiency
- 17:44Resource usage metrics
- 19:08Cleanup and deletion
- 20:33Summary and Q&A
related talks
22mUse Karpenter to scale, optimize & upgrade K8s
27mHarness the power of Karpenter to scale, optimize & upgrade K8s
18mKarpenter : Your trusted Companion for Kubernetes Scaling
36mGroupless Autoscaling with Karpenter
19mJust in time Worker Nodes with Karpenter
01h 01mJust-in-time Nodes for Any AWS EKS Cluster: Karpenter
34mReduce Costs and Optimize on Your Cloud with Karpenter
40mOptimize EKS Cost and Scaling with Karpenter