Skip to content

Remove unusable legacy ND112/ND120 A100 GPU SKUs #1852

Description

Summary

Karpenter still recognizes four legacy A100 GPU SKU names that AKS RP removed from its GPU support map after confirmation that they are unusable:

  • standard_nd112asr_a100_v4
  • standard_nd120asr_a100_v4
  • standard_nd112amsr_a100_v4
  • standard_nd120amsr_a100_v4

Context

PR #1841 synchronized pkg/utils/supported-gpus.yaml with the current AKS GPU matrix while adding Azure Container Linux support. That synchronization removed these entries, but the removal was restored because it was unrelated to ACL and changed existing Karpenter behavior.

Karpenter should evaluate removing these SKUs in a separate, explicitly scoped compatibility change. The current entries must not gain Azure Container Linux support.

Work

  • Confirm no supported Karpenter users or Azure regions still expose these SKU names.
  • Reconcile stale references in Karpenter and upstream GPU catalogs.
  • Remove the four entries from pkg/utils/supported-gpus.yaml if compatibility impact is acceptable.
  • Add tests documenting the intended handling of retired or unusable GPU SKU names.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions