Hard Coded Group name for k3s servers

[fkonradmain]

As already mentioned in #321 , some variable names are not ideal and should be prefixed.

4. token and other parameters should be 'k3s_' prefixed

The host group names for "server" and "agent" are not ideal either. The name is ambiguous. Additionally it is a host variable that is also active on other plays in the same playbook. It could possibly cause collisions with them.

Hence, in my playbook I renamed the server group to "k3s_server" and the agent group "k3s_agent". While this works for the agent playbook, it does not work for the server playbook. Unfortunately, the group names are hard coded into the latter.

This issue comes up, every time the following call is made:

groups['server']

My idea would be to introduce a variable, e.g. inventory_server_group_name and have that name default to server. In that case we would just have to change every occurence of groups['server'] to:

groups['{{ inventory_server_group_name }}']

I am willing to implement this solution myself and provide it as pull request, yet I would like to hear the admin's opinion first.

[dereknola]

Hi @fkonradmain, I understand your issue, but could use your help better understanding "how" you and potentially other users utilize this playbook in concert with multiple playbooks or large inventories. My day to day use of large scale ansible playbooks is limited. I typically am using only 1 or 2 separate playbooks on an small inventory (less than 10 machines).

Because of my use case and potentially other beginners, when I took over as maintainer, I didn't see a reason change all the vars and the group names and add k3s_ it seemed redundant to me. You are using the k3s-ansible playbook, of course all the stuff in the inventory is for K3s.

In the K3s context, server and agent are the official names we give node roles. The naming inside the K3s space is not ambiguous, but very clear and used throughout the k3s-io org and docs.

I'm trying to balance the clarity and ease-of-use for first time ansible users (K3s is so easy to deploy that we often get people with no experience in Kubernetes starting with K3s) and people who may be utilizing this playbook in large scale enterprise deployments.

What does a typically inventory look like for you? Why is server and agent to ambiguous? What other playbooks do you utilize that come in conflict with this one? Would the groups being changed to k3s_server and k3s_agent be enough to resolve your issue?

[simagick]

Currently the server role will try to join servers with the first node in the server group to create the control plane. This means that we can not manage multiple clusters within the same inventory with the hard-coded group names.

Making configurable groups gives us more flexibility. I have a PR that addresses this issue in #331