Add specific variables in order to set kubelet arguments: useful or not?

[sambonbonne]

While implementing a pull request to add custom k3s agent command arguments (see #862), I saw it would be easy to implement new variables (I implemented it on my side but I did not test it for now, I want to know if these variables are useful):

• adding a kubelet_args variable to set common kubelet arguments for control planes and agents nodes
• adding a control_planes_kubelet_args variable to set kubelet arguments for control planes nodes only
• adding a agents_kubelet_args which variables to set kubelet arguments for agents nodes only

With the implementation I tried, these arguments are set directly in generated k3s configuration files (instead of passing arguments to k3s server … and k3s agent ….

My question is: would it be useful or does this idea conflict with k3s_exec_server_args (and the potential k3s_exec_agent_args variable I propose in #862)?

[mysticaltech]

@sambonbonne Because of the behavior layed out below. I think k3s_exec_server_args and k3s_exec_agent_args are enough. What do you think?

[sambonbonne]

@mysticaltech hey I fear this would be too repetitive.

My implementation adds the kubelet_args in the config.yaml and the idea behind this idea is that it would be more "beginner friendly" to have simple variables for this specific purpose, but if think like this, we would have a specific variable for every setting in the k3s config file and it will be a lot of burden.

[mysticaltech]

Yes exactly, I mean, customizing k3s is not "beginner friendly" in any shape or form. So I think we are good as is and k3s_exec_server_args and k3s_exec_agent_args are enough, as you can do everything already.