Service: Remove default values for controller.service.nodePorts & controller.service.internal.nodePorts.

[Gacko]

Tests on workload clusters (not always required)

For changes in the chart, chart templates, and ingress controller container images, I executed the following tests
to verify them working in live enviromnents:

Test / Provider	AWS	Azure	KVM
Upgrade from previous version	✓	✓
Existing Ingress resources are reconciled correctly	✓	✓
Fresh install	✓	✓
Fresh Ingress resources are reconciled correctly	✓	✓

Testing was done using hello-world-app.

Hint for KVM:

kubectl port-forward -n kube-system svc/nginx-ingress-controller-app 8080:80
ingress_domain=host.configured.in.ingress; curl --connect-to "$ingress_domain:80:127.0.0.1:8080" "http://$ingress_domain" -v

[Gacko]

@glitchcrab I could need some review from a KVM / legacy product expert. Basically we'd like to align the $setNodePorts part here to how upstream is doing it.

To not break existing LoadBalancer services in cloud environments, we need to reset the default values for controller.service.nodePorts & controller.service.internal.nodePorts to "". Otherwise cloud LoadBalancer services would get configured to using those node ports. Aside from probable conflicts with other services, this means a service outage. Both bad.

Anyway: Do you think we could ask customers to either set those ports on their own or simply discontinue KVM / legacy support with the upcoming major release?

Another option would be setting those values in the chart values already provided by cluster-operator, but this means releasing a new version of the latter.

[ubergesundheit]

👍