Quickstart Started Guide Does Not Work

[danehans]

What happened:
Followed the steps in the quickstart guide and EPP failed to run as expected due to:

2025-02-26T02:50:28Z    ERROR    controller-runtime.source.EventHandler    source/kind.go:71    if kind is a CRD, it should be installed before calling Start    {"kind": "InferenceModel.infer │
│ ence.networking.x-k8s.io", "error": "no matches for kind \"InferenceModel\" in version \"inference.networking.x-k8s.io/v1alpha2\""}                                                             │

What you expected to happen:
EPP to run and reconcile CRDs.

How to reproduce it (as minimally and precisely as possible):
Follow the guide.

Anything else we need to know?:

The issue is that the quickstart steps are based on the main branch except for the CRD installation that is still using v0.1.0. Additionally, now that the quick moved to site-src the release tooling (release-quickstart) to manage the guide is broken.

Environment:

• Kubernetes version (use kubectl version):
• Inference extension version (use git describe --tags --dirty --always):
• Cloud provider or hardware configuration:
• Install tools:
• Others:

[nirrozenbaum]

/assign

[nirrozenbaum]

moving task to @kfswain

[kfswain]

/assign

Looks like the issue is just running make install since v1alpha2 got pushed recently and the epp manifest uses the main image (which now supports the new api)

[danehans]

We can update the guide with ^ as a temp fix but the guide should be based on a tagged release to ensure stability. We'll need to fix https://github.com/kubernetes-sigs/gateway-api-inference-extension/blob/main/hack/release-quickstart.sh to use the new site-src guide and verify each command uses either the tagged release assets or tagged repo URL, e.g. kubectl apply -f https://github.com/kubernetes-sigs/gateway-api-inference-extension/blob/v0.1.0/pkg/manifests/vllm/deployment.yaml instead of kubectl apply -f https://github.com/kubernetes-sigs/gateway-api-inference-extension/raw/main/config/manifests/vllm/deployment.yaml.

[kfswain]

Fixed by: #410

We can make a version-locked variant as well, but that can be a follow up issue

[danehans]

Here is a quickstart that works for the v0.1.0 release.

[ahg-g]

This is addressed now? can we close it?