Clarification on behavior for duplicate parent refs for routes.

[jm96441n]

What would you like to be added:
An additional conformance test specifying the behavior when a gateway has only a single listener and an http route specifies the gateway twice: once in specifying the gateway with the listener and once without specifying the specific listener. According to the docs this should result in an invalid route configuration.

This feels like a potentially uncovered edge case in the docs as it is neither two identical parents (as could be detected by the webhook) nor two distinct sections, and might be a valid configuration if the Gateway had a second listener which could match the parentRef not specifying a listener name.

Why this is needed:
To clarify the above behavior for implementers of the gateway spec.

[jm96441n]

I wouldn't mind writing the tests themselves, more looking for clarification on what the expected behavior is here.

[shaneutt]

I wouldn't mind writing the tests themselves, more looking for clarification on what the expected behavior is here.

Awesome, appreciate the interest! For extra clarity, can you provide an example YAML configuration of the currently ambiguous situation?

[jm96441n]

hey @shaneutt sorry this took a little bit to get back to you on, below is a yaml configuration that shows the ambiguous situation:

apiVersion: gateway.networking.k8s.io/v1beta1
kind: Gateway
metadata:
  name: apigw
spec:
  gatewayClassName: api-gateway
  listeners:
  - name: http-listener
    protocol: HTTP
    port: 30002
    hostname: '*.example.com'
    allowedRoutes:
      namespaces:
        from: All
---
apiVersion: gateway.networking.k8s.io/v1beta1
kind: HTTPRoute
metadata:
  name: http-route
spec:
  hostnames:
    - ref.example.com
  parentRefs:
    - group: gateway.networking.k8s.io
      kind: Gateway
      name: apigw
    - group: gateway.networking.k8s.io
      kind: Gateway
      name: apigw
      sectionName: http-listener
  rules:
    - backendRefs:
      - kind: Service
        name: some-service
        port: 80
        weight: 1
      matches:
      - path:
          type: PathPrefix
          value: /

so this HTTPRoute references the same gateway twice, but specifies the listener in one parent ref and does not specify it in the other ref. This could be a potentially valid situation if the gateway had more than one listener.

[shaneutt]

Thank you for the illustration. Yes this seems like something we would want to specify so let's do this:

/triage accepted

We'll need whoever takes this on to take some time to build consensus among the community about how to resolve it. It would probably be good to sort it out prior to GA:

/kind conformance
/milestone v1.0.0
/help

[k8s-ci-robot]

@shaneutt:
This request has been marked as needing help from a contributor.

Guidelines

Please ensure that the issue body includes answers to the following questions:

• Why are we solving this issue?
• To address this issue, are there any code changes? If there are code changes, what needs to be done in the code and what places can the assignee treat as reference points?
• Does this issue have zero to low barrier of entry?
• How can the assignee reach out to you for help?

For more details on the requirements of such an issue, please see here and ensure that they are met.

If this request no longer meets these requirements, the label can be removed
by commenting with the /remove-help command.

In response to this:

Thank you for the illustration. Yes this seems like something we would want to specify so let's do this:

/triage accepted

We'll need whoever takes this on to take some time to build consensus among the community about how to resolve it. It would probably be good to sort it out prior to GA:

/kind conformance
/milestone v1.0.0
/help

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[shaneutt]

@robscott is going to create a follow-up issue for this to create a conformance test to resolve this.

[robscott]

I haven not forgotten about this one, waiting for clarification on #2326 before writing up what we'd expect a conformance test to look like for this.

[shaneutt]

I talked with @sunjayBhatia about this one and he can take this one on.

/assign @sunjayBhatia

We expect resolving #1626 completes most of this.

Thanks @sunjayBhatia !

[sunjayBhatia]

Looks like #2350 covers this issue? I refrained from making any changes regarding this in #2396 since that PR already existed

[youngnick]

Ah sorry, yes it does, I should have taken this issue.

[sunjayBhatia]

No worries! I can assign it over to you since you've got that in progress already (if I have powers to do so)

Generally looks like this has a lot of overlap with #2326 so probably best resolved together?

/assign @youngnick