Question: What is CABP3's expectation of k3s?

[mquhuy]

Hi. I'm pretty new to not just CABP3, but k3s as well, so advance apology if the questions are silly.

What I have been trying to do is to replace kubeadm with k3s in a baremetal setup with metal3 as the infra provider. I notice that I could only make k3s work inside the provisioned machine if I had this in the preK3sCommands block:

curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="server" sh -s -

I'm wondering if I am doing something wrong here, because that command should install and start the k3s systemd service, so there's basically nothing else to do in the k3sCommand (what happened between preK3sCommaands and postK3sCommands)? As an experiment I tried installing k3s without starting the service ( K3S_SKIP_START="true), then the k3s service in the provisioned machine would become inactive.

CABPK doesn't require any manual manipulation for kubeadm. Is it so that CABP3 does it differently and expect users to handle k3s service on their own? In that case, do we have to manually put the domain and token in as well?

[mogliang]

You don't need to add prek3sCommands below, it shall already in generated cloudinit userdata.

curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="server" sh -s -

You can find the generated cloudinit secret in mgmt cluster and b64 decode, see if there are anything goes wrong.