Changing Service Subnet in Talos OS Cluster – Possible Risks and Issues

[erelbi]

We are planning to change the service subnet in our Talos OS–based Kubernetes cluster.
Before proceeding, we would like to understand the potential risks, side effects, and required steps.

Questions:

1. What is the recommended way to safely change the service subnet in a Talos cluster?
2. After applying the change, will existing services or kube-proxy configurations break?
3. Do we need to manually restart pods, reconfigure kube-proxy, or recreate services/endpoints?
4. Are there any known issues with DNS resolution (CoreDNS) or cluster networking after such a change?

We want to ensure minimal downtime and avoid unexpected disruptions. Any guidance, best practices, or known caveats would be greatly appreciated.

[smira]

This has nothing to do with Talos Linux itself, as it doesn't do any Kubernetes CNI or kube-proxy handling.

Talos Linux only pushes service subnet from the machine configuration to the kube-controller-manager.

All answers to your questions depend on the CNI you are using.

I would not expect in general a change to service subnet to be downtime-free. I guess you will need to apply the change to all machines, restart kube-proxy (if you're using it), and delete and re-create all services to get them new service IPs.

I don't know if it can be done without downtime, unless there is a specific route in some specific CNI to do this.

At this point I would say it might be easier to re-create the clsuter with new settings.