How to safely upgrade control plane node server type and apply kubelet-args?

[asemarian]

I'm currently running a production cluster using the kube-hetzner project, with three control-plane nodes and six worker nodes. Recently, I encountered an issue (detailed here) that was causing worker nodes to go down. After some research, I’ve decided to take the following steps to address the problem:

1. Upgrade control plane node server types
I plan to upgrade the server type for my control-plane nodes from CPX11 to CPX21. I am not sure what the recommended approach is:

• Should I just update the server_type key and apply the changes?
• Or do I set the count to 0 and create a new nodepool?
• In case of the second approach, can I safely apply it to the first control-plane nodepool?

2. Add global kubelet-args
Specifically, I want to add these three:

• --kube-reserved="cpu=100m,memory=600Mi,ephemeral-storage=1Gi,pid='1024'"
• --eviction-hard="imagefs.available<15%,memory.available<100Mi,nodefs.available<10%,pid.available<10%"
• --enforce-node-allocatable=pods

Since these flags need to be applied to all nodes in the cluster, I am concerned about applying this change without testing it.

• When I make this change, does it require the nodes to be drained?
• Do I need to reboot the nodes?
• Do you have recommendations for how to test this safely before applying it?

Any guidance regarding this would be greatly appreciated 🙏

[asemarian]

Turns out the server type can be changed in-place. As for adding kubelet args, it's possible to do that globally for all nodes in the cluster or set them for each node separately.