aws-eks: neuron device plugin manifest better reference
- Dominant language
- TypeScript
- Stars
- 12.9k
- Forks
- 4.6k
- Avg merge
- 1d 19h
- Merged PRs (30d)
- 74
Description
### Describe the bug
the neuron device plugin addon used in the cdk uses a custom manifest, see here: https://github.com/aws/aws-cdk/blob/f3d74bb78189ec6b76cfa85c97d993c1b26c1cac/packages/aws-cdk-lib/aws-eks/lib/cluster.ts#L1979
which is NOT pointing to the official neuron image (public.ecr.aws/neuron/neuron-device-plugin)
and rbac is missing
going into crashloopback and preventing metrics to be exposed
### Expected Behavior
the right files are used
### Current Behavior
crashloopback on deployment of inf2.xlarge
### Reproduction Steps
deploy on inf2
### Possible Solution
the neuron device plugin addon used in the cdk uses a custom manifest, see here: https://github.com/aws/aws-cdk/blob/f3d74bb78189ec6b76cfa85c97d993c1b26c1cac/packages/aws-cdk-lib/aws-eks/lib/cluster.ts#L1979
while there is a better existing reference from the Neuron, see description here: https://awsdocs-neuron.readthedocs-hosted.com/en/latest/containers/tutorials/k8s-setup.html
the yaml to use is https://raw.githubusercontent.com/aws-neuron/aws-neuron-sdk/master/src/k8/k8s-neuron-device-plugin.yml
and also rbac needs to be used which is not in the current implementation
const RBAC_URL = "https://raw.githubusercontent.com/aws-neuron/aws-neuron-sdk/master/src/k8/k8s-neuron-device-plugin-rbac.yml
### Additional Information/Context
_No response_
### CDK CLI Version
2.130.0
### Framework Version
_No response_
### Node.js Version
v20.4.0
### OS
sonoma 14.3
### Language
TypeScript
### Language Version
_No response_
### Other information
_No response_
Contributor guide
Research direction
Start at packages/aws-cdk-lib/aws-eks/lib/cluster.ts around the manifest referenced near line 1979. Compare the current custom manifest with aws-neuron-sdk/src/k8/k8s-neuron-device-plugin.yml and k8s-neuron-device-plugin-rbac.yml, then reproduce on an inf2.xlarge cluster. Done means the addon deploys without a crash loop and exposes the expected metrics.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- aws, kubernetes, typescript
- Domain
- cloud, infrastructure
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100