Handling Node Churn in Karpenter: Efficient Scaling for Large EKS... Shivani Mehrotra & Chetan Saini

Learn how to effectively manage node churn in large Amazon EKS clusters using Karpenter, an open-source Kubernetes autoscaler, in this 27-minute conference talk from CNCF. Discover practical strategies for optimizing node provisioning and minimizing churn across multiple EKS clusters when dealing with fluctuating workloads that can lead to instability and high costs. Explore key techniques including Pod Priority implementation, Pod Disruption Budgets (PDBs), and node pool configuration to improve scheduling efficiency and reduce resource waste. Understand how to reduce do-not-disrupt annotations, tighten affinity rules, and align Karpenter nodes with EC2 savings plans while leveraging Karpenter's latest features like reserved instances for enhanced performance and cost savings. Gain insights into understanding the impact of node churning on both costs and cluster stability, learn optimization techniques for efficient node scaling with Karpenter, and discover methods to achieve significant cost savings through spot instances and refined scaling policies. Master the implementation of strategies that make EKS clusters more efficient and cost-effective while maintaining stability in dynamic workload environments.