~/ nmadiraju.com Amazon EKS · Karpenter NodePools and disruption
arrow_back Back to nmadiraju
Visual Walkthrough · Amazon EKS compute

How Karpenter NodePools and Disruption Work on Amazon EKS

Follow one team's general, gpu and batch NodePools from a catch-all to bounded pools, then through drift, expiry and consolidation, and check yourself with 8 questions.

1Visual Walkthrough

Karpenter NodePools on EKS: four NodePools, an EC2NodeClass and the nodes each one launched On the left, the platform team applies NodePools and the EC2NodeClass, which holds the AMI alias; Karpenter reads both. Across the top are the NodePool cards: a catch-all default at first, then general-sp, general, gpu and batch with their requirements, limits and disruption settings. Below, inside your VPC, are the EC2 nodes each NodePool launched. The walkthrough shows GPU nodes shared by batch, an unlimited NodePool, a weighted NodePool falling back, AMI drift, expiry, consolidation blocked by a scheduled budget, and a do-not-disrupt Pod. Platform team kubectl apply EC2NodeClass default · AMI alias: subnets, SGs: by tag role: KarpenterNodeRole Karpenter kube-system YOUR VPC · us-east-1 TERMINATED TERMINATED TERMINATED TERMINATED TERMINATED TERMINATED TERMINATED TERMINATED TERMINATED TERMINATED

    2Quiz to check understanding

    Check yourself