aws / aws/amazon-vpc-resource-controller-k8s
Use dynamic instance lists in VPC resource controller
- Dominant language
- Go
- Stars
- 106
- Forks
- 71
- Avg merge
- 10h 28m
- Merged PRs (30d)
- 6
Description
**Is your feature request related to a problem? Please describe.**
When AWS releases a new instance type there is a 30 day delay before Karpenter users are able to test on the new instance types. An internal AWS API is a bottleneck for EKS Optimized AMI, Karpenter, and the vpc-resource controller.
**Describe the solution you'd like**
I would like the ability to test GA instance types in Karpenter on Day 1 in supported regions, with the ability to discover all instance specifications and limits to consider for configuration planning.
**Describe alternatives you've considered**
Alternative approaches require engagement with AWS service team to request for expedited support or version updates, or to run Codegen across multiple repos to populate missing values with public API.
**Additional context**
Pod Security Groups (SGPP) require existence of values populated in the instance lists so they take a dependency on the internal AWS API.
Contributor guide
Research direction
Start by tracing how the VPC resource controller obtains its instance lists and how Pod Security Groups (SGPP) consume those values. Compare the current internal AWS API dependency with the public API and Codegen alternatives; done means supported regions can discover GA instance specifications and limits on day one without waiting for internal list updates.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- aws, go
- Domain
- networking
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100