Hi all, I have a special case. I have an rancher manager and RKE2 cluster with 3 nodes. The RKE2 cluster was made with a custom provider, which means I created 3 new Rocky Linux machines and then did the registration with generated curl command. My problem is when I would like to shrink the disk of the machines or change the underlying OSS, I need to regenerate the nodes. So I drain the node and delete them, then I create a new host and register it with the curl command. This is all OK until I come to the first controller node, which is the ETCD bootstrap nodes. When I remove the bootstrap node and switch the ETCD leader, when I add a new worker node, I don't get the right bootstrap IP address for the bootstrap service and can't register the new node to the rancher manager. Is there any suggestion on how to fix this problem or make any workaround to change or set the IP address when generating the curl command? Thank you for your awareness.

In my experience, I've built external Nginx as proxy server to all of the cluster nodes, and make the SSL termination in the Nginx. I set the registration address to the vhost name of the nginx. So, full secure of the connection with domain name, no need to change the registration IP, just use the domain that handled by the reverse proxy. Even better, use redundant Nginx using corosync to build HACluster no single point of failure

Your are instaling RKE2 befor importing it to rancher manager? In this case i would to rke2 + kube-vip and use tke VIP address for the server join uri. But i am stuck with the registered RKE2.

In my conditions, just OS then using the registration link to add the cluster. Is the VIP changes? I think better use domain that point to the VIP, and it could use DNS record, also check the registration URL in the RAncher manager, is the same with VIP address or not