Learn how Karpenter provisions and consolidates cluster capacity by using NodePools, NodeClasses, and NodeClaims, plus how it compares to Cluster Autoscaler.

DataDog Blog offers insights, tutorials, and updates on monitoring, analytics, and observability solutions. Covering topics such as infrastructure monitoring, log management, and application performance monitoring, DataDog Blog provides resources for developers, DevOps engineers, and IT professionals. Developers can learn about best practices, troubleshooting techniques, and optimization strategies for managing complex systems through DataDog's blog posts and guides.

Datadog

Karpenter is a Kubernetes autoscaler that provisions right-sized compute directly via cloud provider APIs, replacing the traditional Cluster Autoscaler (CAS). Its configuration model separates provider-agnostic NodePools (defining constraints like instance families, capacity types, and consolidation policies) from provider-specific NodeClasses (defining AMIs, subnets, security groups). Karpenter batches unschedulable pods to provision fewer, better-utilized nodes just-in-time, tracks in-flight NodeClaims to avoid duplicate provisioning, and continuously consolidates workloads through emptiness, underutilization, and drift detection. Compared to CAS, Karpenter offers faster provisioning via direct API calls, flexible instance selection beyond static node groups, and proactive cost optimization through workload consolidation rather than just deleting empty nodes.

Understanding Karpenter architecture for Kubernetes autoscaling

How Karpenter optimizes performance and cost

Comparing Karpenter to Cluster Autoscaler

Building a foundation for Karpenter observability