10/20/2022, 8:07 PM
I am having some trouble with a Rancher-managed RKE(1) cluster running on Harvester, provisioning fails to add new etcd-role members because it first tries to remove members that no longer exist (it already succeeded in removing them, but doesn’t seem to have noticed this, and just keeps trying). I can’t figure out where the state is that remembers that these members exist (they are no longer in the Rancher cluster as Machines or references in the ManagmentCluster, they are not in the managed cluster’s etcd, VMs are not in Harvester) such that it is trying to remove them.
Given the nature of my issue, I’m not sure which channel would be the most appropriate to seek help in.
Reading rancher code and RKE code, It looks like RKE is in charge of this reconciliation so I’ll address that channel.