Trying to restore my broken etcd of a cluster managed in Rancher the GUI gives me an error message "Cluster must have at least one etcd plane host: failed to connect to the following etcd host(s) ...". How is this supposed to work? AFAICS the documentation of Rancher 2.6 doesn't describe this special case (nor does it give any other example for a successful restore).
12/17/2022, 11:50 AM
Could you provide the full logs? I suspect it is an issue with the private key & you might need to regenerate a pair to reconnect successfully. You can see this issue if the error message from your logs match the ones mentioned here.
If this did solve your issue, please could you let me know? I shall take it back to the RKE/Rancher team as some thing that we need to document.
12/19/2022, 7:24 AM
Yes, my cluster is working again. I deleted the broken etcd container using docker and redeployed the control plane node using the docker command line provided by the Rancher GUI. I would assume the most recent backup of the etcd database was included automatically.
Please note that this is not my recommendation how this problem can be solved in general. Your mileage may vary.
PS: I don't have the log of the broken etcd container anymore, but AFAIK it complained about a broken etcd database. #2448 seems to be about a fresh installation. I doubt that these are similar problems.