DRA: Handle core resource requests through DRA

[pravk03]

Enhancement Description

Using DRA drivers for core resources (like dra-driver-cpu), enables fine-grained resource management but introduces a critical accounting problem. The standard kube-scheduler resource accounting model is unaware of allocations made by the DynamicResources plugin, and vice-versa, leading to a potential resource oversubscription on nodes. While a synchronization model similar to the one used for DRA Extended Resources is needed, the same solution cannot be applied directly. The key difference is that extended resources are advertised either through the node object (node.status.allocatable) or via ResourceSlice, but not both simultaneously for the same node. In contrast, core resources like CPU are always present in node.status.allocatable, and a DRA driver would also manage them via ResourceSlice. This simultaneous representation creates a unique challenge that requires a unified accounting model.

cc: @johnbelamaric, @mortent , @ffromani , @yujuhong , @dchen1107

• One-line enhancement description (can be used as a release note): Handle resource accounting for core resources (CPU, Memory etc.) managed through DRA
• Kubernetes Enhancement Proposal: TBD
• Discussion Link:
  • Shared ResourceClaim Causes Incorrect CPU Accounting kubernetes-sigs/dra-driver-cpu#6
  • Design exploration doc
• PRs by stage and milestone:
  • Alpha - v1.xx
    • KEP (k/enhancements) update PR(s):
    • Code (k/k) update PR(s):
    • Docs (k/website) update PR(s):

Please keep this description up to date. This will help the Enhancement Team to track the evolution of the enhancement efficiently.

[pravk03]

/wg device-management
/sig scheduling
/sig node

[swatisehgal]

/cc

[pohly]

Are there plans to work on this for 1.35? Who wants to own this?