aws / aws/containers-roadmap

[EKS-AutoMode] [request]: Support for NodeOverlay

Open
#2,690 1 comment 10 reactions 0 assignees View on GitHub
EKS EKS Auto Mode Proposed
Dominant language
Shell
Stars
5.4k
Forks
334
PR merge metrics
No merged PRs in 30d

Description

### Community Note

* Please vote on this issue by adding a 👍 [reaction](https://blog.github.com/2016-03-10-add-reactions-to-pull-requests-issues-and-comments/) 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

**Tell us about your request**
It could be very helpful to enable NodeOverlays ([link](https://karpenter.sh/docs/concepts/nodeoverlays/)) in EKS-AutoMode, to allow customization of node price/capacity (e.g., to implement proper auto-scaling with fractional GPUs).

**Which service(s) is this request for?**
EKS Auto Mode

**Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?**
For example, when using NVIDIA GPU fractioning with MIG "mixed" strategy ([docs](https://docs.nvidia.com/datacenter/cloud-native/kubernetes/latest/index.html#the-mixed-strategy)), cluster auto-scaling does not work, because Karpenter does not recognize the nvidia.com/mig-* resources.

**Are you currently working around this issue?**
NodeOverlays is a new feature in Karpeneter designed to address these (and other) use-cases. However, it is not enabled in EKS Auto Mode.

Contributor guide

Open the contributing guide

Research direction

Start with the NodeOverlays documentation linked in the request and compare it with EKS Auto Mode behavior. Review the NVIDIA mixed MIG strategy documentation to understand the autoscaling failure; done means EKS Auto Mode supports NodeOverlay-based customization and the described fractional-GPU capacity case can scale correctly.

Written by the indexing model from the issue text.

Assessment

Tech stack
aws, kubernetes
Domain
cloud, infrastructure
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.