talosctl upgrade-k8s ignores machine/kubelet/clusterDNS configuration

[tlemarchand]

Bug Report

Description

I had to change my kube-dns service IP. I've updated my cluster configuration files to reflect the change, so my cluster is in functional, pods can resolve names.

version: v1alpha1 # Indicates the schema used to decode the contents.
debug: false # Enable verbose logging to the console.
persist: true
# Provides machine specific configuration options.
machine:
    kubelet:
        # # The `ClusterDNS` field is an optional reference to an alternative kubelet clusterDNS ip list.
        clusterDNS:
            - 10.96.0.42

However I cannot upgrade my kubernetes cluster anymore, because upgrade-k8s tries to rewrite my kube-dns service to old values :

talosctl --nodes controlplane01 upgrade-k8s --to v1.34.3 --dry-run

[...]

 < dry run, change skipped
1 error(s) occurred:
        Service "kube-dns" is invalid: spec.clusterIPs[0]: Invalid value: []string{"10.96.0.10", "fd00::a"}: may not change once set

I believe it's caused by this function :

talos/pkg/machinery/config/types/v1alpha1/v1alpha1_clusterconfig.go

Line 264 in 150c41c

return sideronet.NthIPInCIDRSet(serviceCIDRs, 10)

In my opinion, the machine/kubelet configuration, if it exists, should be used to generate kube-dns service. This way any service IP could be used for kube-dns and upgrade would still be possible.

Logs

Environment

• Talos version: v1.11.6
• Kubernetes version: v1.33.4
• Platform: "bare metal" (actually VMs in Proxmox)