Karpenter v1.15: GPU-aware provisioning, smarter node repair, and more
Containers Blog
This article announces Karpenter v1.15, a major update to the Kubernetes node lifecycle manager that adds hardware-aware provisioning, improved repair mechanisms, and consolidation enhancements.
- GPU and EFA topology-aware provisioning via Dynamic Resource Allocation (DRA) for peer-to-peer device communication
- Volume-limit-aware provisioning prevents unnecessary node launches for workloads with many PersistentVolumeClaims
- Nitro Enclaves support enables confidential computing workloads on Karpenter-managed nodes
- Node repair now respects Pod Disruption Budgets and NodePool disruption budgets as voluntary disruption
- Reboot-as-repair action clears transient GPU faults without full instance replacement
- Pod deletion cost management aligns Karpenter consolidation with Kubernetes ReplicaSet scaling decisions
- CEL expressions in kubelet configuration enable dynamic per-instance-type settings in single EC2NodeClass
Karpenter v1.15 bridges the gap between Kubernetes abstractions and physical hardware requirements, improving provisioning efficiency and workload availability for distributed training, stateful services, and mixed-instance fleets.
The AWS News Feed is currently looking for gold sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.
The AWS News Feed is currently looking for silver sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.