Support hugepages-2Mi resource type

[DanielJuravski]

Tell us about your request

Support hugepages-2Mi resource type.

Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?

I need to apply various workloads that should allocate different values for hugepages-2Mi resource type.

resources:
  limits:
    habana.ai/gaudi: 8
    hugepages-2Mi: "21000Mi"
    memory: 50Gi
  requests:
    habana.ai/gaudi: 8
    hugepages-2Mi: "21000Mi"
    memory: 50Gi

The main instance type I'm using is dl1.24xlarge - it doesn't support hugepages-2Mi resource type https://karpenter.sh/v0.22.1/concepts/instance-types/#resources-148 (none of the Karpenter's instances types do).
Thus I can't apply workload with custom hugepages-2Mi value.

Are you currently working around this issue?

No

Additional Context

The message from Karpenter log when it does not work:

Failed to schedule pod, incompatible with provisioner "gaudi", no instance type satisfied resources
{"cpu":"24","_[_habana.ai/gaudi":"8","hugepages-2Mi":"21000Mi","pods":"1_|http://habana.ai/gaudi":"8","hugepages-2Mi":"21000Mi","pods":"1]_"}
and requirements_ [kubernetes.io/os](http://kubernetes.io/os) In [linux], karpenter.sh/capacity-type In [on-demand], [node.kubernetes.io/instance-type](http://node.kubernetes.io/instance-type) In [dl1.24xlarge], karpenter.k8s.aws/instance-family In [dl1], [habana.ai/gaudi](http://habana.ai/gaudi) In [true], karpenter.sh/provisioner-name In [gaudi], [kubernetes.io/arch](http://kubernetes.io/arch) In [amd64]

Attachments

No response

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

[ellistarn]

I'm not too familiar with huge pages. Is this set for all nodes, or only under some settings?

[DanielJuravski]

All nodes,

For Linux workloads, you can specify huge page resources. Huge pages are a Linux-specific feature where the node kernel allocates blocks of memory that are much larger than the default page size.

https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#resource-types
https://aws.amazon.com/premiumsupport/knowledge-center/configure-hugepages-ec2-linux-instance/

[jonathan-innis]

We're tracking extended resources as part of kubernetes-sigs/karpenter#751. I'd recommend +1-ing that issue. Closing this as a duplicate.