Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Help improve this page

To contribute to this user guide, choose the Edit this page on GitHub link that is located in the right pane of every page.

Control deployment of workloads into EC2 On-Demand Capacity Reservations with EKS Auto Mode

EC2 On-Demand Capacity Reservations (ODCRs) allow you to reserve compute capacity for your Amazon EC2 instances in a specific Availability Zone for any duration. When using EKS Auto Mode, you may want to control whether your Kubernetes workloads are deployed onto these reserved instances to maximize utilization of pre-purchased capacity or to ensure critical workloads have access to guaranteed resources.

By default, EKS Auto Mode automatically launches into open ODCRs. However, by configuring capacityReservationSelectorTerms on a NodeClass, you can explicitly control which ODCRs your workloads use. Nodes provisioned using configured ODCRs will have karpenter.sh/capacity-type: reserved and will be prioritized over on-demand and spot. Once this feature is enabled, EKS Auto Mode will no longer automatically use open ODCRs—they must be explicitly selected by a NodeClass, giving you precise control over capacity reservation usage across your cluster.

Example NodeClass

apiVersion: eks.amazonaws.com/v1
kind: NodeClass
spec:
  # Optional: Selects upon on-demand capacity reservations and capacity blocks
  # for EKS Auto Mode to prioritize.
  capacityReservationSelectorTerms:
    - id: cr-56fac701cc1951b03
    # Alternative Approaches
    - tags:
        app: "my-app"
      # Optional owning account ID filter
      owner: "012345678901"

This example NodeClass demonstrates two approaches for selecting ODCRs. The first method directly references a specific ODCR by its ID (cr-56fac701cc1951b03). The second method uses tag-based selection, targeting ODCRs with the tag Name: "targeted-odcr". You can also optionally filter by the Amazon account that owns the reservation, which is particularly useful in cross-account scenarios or when working with shared capacity reservations.