Description

Observed Behaviour:

If nodes are not in ready state, an infinite number of nodeclaims can be created.

Expected Behaviour:

Nodeclaims cannot be created if the sum of nodeclaim resources are greater than the nodepool limits.*

*within reason, as there are multiple async threads provisioning an extra "surge" of a few nodeclaims is expected.

Reproduction Steps (Please include YAML):

1. Edit nodeconfig in such a manner that nodes cannot join the cluster. eg for AWS you can set the osFamily to AL2 and the AMI to AL2023 If I recall correctly.
2. Create a pod
3. Nodeclaim is created
4. Pod is deleted
5. Nodeclaim remains
6. GOTO 2.

In my case, we had defined CPU Limits on the nodepools.

Versions:

• Chart Version: Karpenter 1.3
• Kubernetes Version (kubectl version): 1.31

• 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