Improve documentation for publishedService and IP options

[samip5]

What does this PR do?

Makes it more clear what the kubernetes-ingress options actually do. like ip and publishedService.

Motivation

Hackaeton and the long-lasting issue #8406.

Fixes #8406

More

• Added/updated documentation

[bwbrock]

These changes appear good.

[kevinpollet]

Thanks 👍

[rtribotte]

Thanks 👍

[samip5]

Hi @kevinpollet. I'm little sad that you modified the issue description to add the "fixes" as this doesn't in reality fix the underlying bug which is in using Endpoints instead of EndpointSlices. The bug itself is still in the code.

[kevinpollet]

Hello @samip5, and sorry for that, I missed something.
Could you please open a detailed issue, describing the bug with a reproduction case?

[samip5]

Hello @samip5, and sorry for that, I missed something. Could you please open a detailed issue, describing the bug with a reproduction case?

It's right there #8406 (comment) and has been for multiple years.

[kevinpollet]

Hello @samip5,

If I am not mistaken, issue #8406 was about copying the LoadBalancer service IPs to the Ingress status. This is working as expected by using the publishedService option and confirmed in the following comment (#8406 (comment)). This is why I closed the issue, after the clarification you made in the documentation.

The following comment #8406 (comment) is about the IPs used for load balancing, which seems to contain only one IP Family. If there is an issue with IPv6 or dual stack clusters, it would be easier to open a new issue to be able to address it properly.

Could you please open this new issue describing the problem?

[samip5]

Hello @samip5,

If I am not mistaken, issue #8406 was about copying the LoadBalancer service IPs to the Ingress status. This is working as expected by using the publishedService option and confirmed in the following comment (#8406 (comment)). This is why I closed the issue, after the clarification you made in the documentation.

The following comment #8406 (comment) is about the IPs used for load balancing, which seems to contain only one IP Family. If there is an issue with IPv6 or dual stack clusters, it would be easier to open a new issue to be able to address it properly.

Could you please open this new issue describing the problem?

You're unfortunately mistaken, as the original one was also about EndpointSlices (currently the copying only copies the first IP family due to the issue of not using EndpointSlices aka #8406 (comment)), not just copying the status thus it shouldn't have been closed when the PR was merged for docs.

[emilevauge]

Hi @samip5
I propose that you open a specific issue to explain your problem.
Thank you

[samip5]

Hi @samip5 I propose that you open a specific issue to explain your problem. Thank you

It already has been explained (multiple times, mind you, even in the Hackathon), but no matter how I try to explain it, it seems maintainers will not understand my point, so I will rather just not use Traefik.

Maybe @aojea wants to explain the problem, but I have already tried to no avail.

[emilevauge]

I'm sorry, but that's how most open source project works: to report problems through issues.
This is the only way to avoid confusion and understand your use case (which is not the case today).
Don't hesitate to come back later if you want to report bugs or contribute :)