-
Notifications
You must be signed in to change notification settings - Fork 485
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
v1alpha2 feedback #790
Comments
Thanks @howardjohn! This is a great list, I've added this issue to the v1alpha2 milestone. |
Looks like this got out of sync, but there was a relatively last minute change to make this field required. I missed updating the docs to correspond with that change. |
I think it should be ignored. Each implementation can't know what is or isn't valid for other implementations.
If a controller can tell that a Route is trying to attach to a Gateway but not allowed to, I think that should be communicated in Route status at a minimum, maybe even Gateway status. |
Awesome list, thanks @howardjohn. A couple thoughts:
Good catch, I think it should be something like:
In the above, I'm using "ensured" to mean "must be present and". (I have on my list to do a pass over the status stuff to clarify what Conditions should be present and when, tl;dr I think all Conditions should always be present, but that's a discussion for another issue).
I think this is a good point, and that if we are going to have
Yeah, it's really a mutli-value key of port, protocol, TLS settings. I think it's probably more correct to say "for the specified Listener", but that makes it a little redundant.
I agree that Passthrough TLS should only be supported on TLSRoute objects, under a "TLS" protocol listener. Otherwise, it implies a level of visibility into the HTTP stream that TLS passthrough doesn't allow.
I think that we should also strongly recommend that people pull down their resources, remove v1alpha1, and then install v1alpha2 as the upgrade flow. I don't think that it's a good idea to have both versions in a cluster at once because of things like the above. |
We already set the default to Namespace.
I don't think we can even if we wanted to. So no.
If you (Gateway) are referenced in the ParentRefs, then it is your responsibility to report the status. |
@howardjohn / @robscott can we split these checkboxes off into issues? Github has a feature to do that and will let you convert a task to an issue, replacing the current text with a link to the new issue. I know its a lot of issues to create but from my perspective being able to just assign a couple issues to myself and get them done helps me organizationally. |
Closing this since the only remaining issue is tracked by #838 /close |
@robscott: Closing this issue. In response to this:
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. |
I started working on implementing v1alpha2 and came across a few minor things that I think could use tweaks. Opening a single issue to capture them, and may use as a running log if I have any others.. we can split out larger ones if needed
kubectl get gateway
. We should have the kustomize only deploy v1alpha2 to the cluster. PR Versioning CRD yaml to simplify only installing v1alpha2 #789hostnames
like HTTPRoute, it should IMO also have the same logic described for HTTPRoute there as well.When empty, the "core" API group is inferred.
. But its required. so are users supposed to dogroup: ""
? Isgroup: "core"
legal? EDIT: read examples, looks likegroup: ""
is the expectation. A bit odd but ok. I feel like default of core may be a good idea here - the most common usage will be Service and Secret probably (and routes I guess)The text was updated successfully, but these errors were encountered: