This message was deleted.

isn't the simplest way is to: • remove CP node from etcd/cluster • clean config/runtime data • add a brand new worker using the same machine? What is the reason for that? Maybe you would just like to add tolerations for some workloads to schedule them on CP node?

as a rookie, 3years ago, I created 6 nodes cluster, all rke2-server. Still running in production, but I wanted to convert 3 nodes to agent and keep only 3 nodes as CP. I know, I can readd the node (and noticed, that currently rke seems to correctly remove etcd member, if a node is removed from the cluster). Just wondered, it there is more simple way like restarting rke2 with correct type.

You need to reinstall rke2-server to rke2-agent, I think mostly configurations can remain (recheck labels & tolerations in config), though if you have manifests in server directory they aren't needed anymore, but you can delete and optionally backup server dir also remove server node first, and then re-join node as a worker I think these are the minimal steps (as Michal wrote)

ok, so no other option than removing and then joining node. got it.