v1 Laundry List for AWS

[njtran]

Description

What problem are you trying to solve?
This issue contains a list of breaking API changes that we want to make for v1 for AWS APIs.

• 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

[jonathan-innis]

We should look at the karpenter.k8s.aws/instance-memory and karpenter.k8s.aws/instance-cpu metrics to see if they are still reasonable to use. I can see these labels being a little clunky since they don't necessarily always map to the allocatable and different DS scheduling may give you different available resources for workloads.

[garvinp-stripe]

We have also noticed that EC2's hypervisor takes some random amount of memory (7-5% depending on instance type), so we usually don't rely on any pre-determined labels for node level resources at run time.

[jonathan-innis]

We should investigate dropping the karpenter.sh/managed-by tag for instance resources. This tag was added as a migration mechanism when Machines were introduced, but kubernetes.io/cluster/<cluster-name> basically gives us the same information as the karpenter.sh/managed-by tag does today.

[jonathan-innis]

We should settle on what we want our EC2NodeClass stable JSON printer columns to be.

[jmdeal]

We should drop this bit of ENI configuration magic, given that #5437 merges before v1 (I have no reason to believe it wouldn't). This was originally introduced to solve #3815 but #5437 introduces the long term solution discussed there.

[jmdeal]

We should decouple the release process for the AWS provider and the upstream library: #5765

[jonathan-innis]

Closing out since #6604 was closed.