Feature: Add SMT/hyperthreading control for AKSNodeClass via VMSizeProperties

[fabioaraujopt]

Summary

Azure Karpenter currently does not expose a first-class API to control SMT/hyperthreading behavior for nodes provisioned from AKSNodeClass.

AWS users can influence SMT behavior in their provider workflows, but Azure AKSNodeClass has no equivalent provider-native setting, creating a cross-provider capability gap.

Problem

For workloads where SMT policy matters, the current Azure provider behavior is too rigid:

• No explicit way to request 1 vCPU per physical core (SMT disabled)
• No provider-level API field for users to express this intent in AKSNodeClass
• Teams must accept defaults or rely on non-ideal workarounds

Impact

Lack of SMT controls affects users that need:

• Performance consistency for CAE/HPC-like workloads
• Security hardening scenarios where disabling SMT is part of mitigation policy
• Provider parity between AWS and Azure Karpenter experiences

Proposed Solution

Add support for Azure VM vCore customization in AKSNodeClass:

• Add vmSizeProperties to AKSNodeClassSpec
• Support vmSizeProperties.vCPUsPerCore with validation range 1-2
• Apply the value to VM hardwareProfile at VM creation time
• Regenerate CRDs and include tests for validation + conversion/mapping logic

Constraints / Notes

• Azure requires support at VM size/image/API level
• Configuration is applied at VM creation time (not mutable on running VMs)
• Expected values:
  • 1 = SMT disabled
  • 2 = SMT enabled/default

References

• PR implementing this feature: feat: Add VMSizeProperties support to disable SMT/hyperthreading #1419
• Azure docs (vCore customization): https://learn.microsoft.com/en-us/azure/virtual-machines/vm-customization
• Azure announcement (disable multithreading): https://techcommunity.microsoft.com/blog/azurecompute/announcing-preview-of-vcore-customization-disable-multithreading--configurable-c/4462417

Outcome

Users can declaratively control SMT policy in Azure through AKSNodeClass, improving performance/security posture and reducing provider capability gaps.