I have a question, or at least a discussion topic, for the topology for using a combination of Rancher and Harvester. Here is the minimal equipment/end goal I am trying to reach:
- 1 “workload” cluster: 3 bare metal servers from Hetzner/OVH/etc with Harvester as the base OS, at least 2 VM’s per server, 1 VM for RKE2 server and 1 VM for RKE2 agent.
- 1 Rancher server cluster: 3 bare metal servers with Harvester or cloud VM’s for a total of 3 RKE2 server VM’s and 3 RKE2 agent VM’s. I will install Rancher server on here.
So what order do I do this in? I will assume I have 6 bare metal servers to keep Harvester relevant for all of this. The best solution I can come up with is:
- Install Harvester on all 6 servers via iso image.
- Create two Harvester clusters with 3 servers each.
- Manually create 2 openSUSE VM’s on each of the first 3 servers via the Harvester UI.
- Install RKE2 server and RKE2 agent in their own VM on the first 3 servers.
- Install Rancher Server manually on each RKE2 agent via Helm.
- Import second Harvester cluster (3 bare metal servers) into Rancher.
- Create second RKE2 cluster via the Harvester node driver.
- Deploy applications to second RKE2 cluster.
My sticking point with the above is having to manually create the openSUSE VM’s on the Rancher server cluster, then manually install RKE2, then manually install Rancher server.
It would be ideal to just install Harvester on all 6 machines, set up some configs for the Rancher server cluster, create it automagically with the Harvester node driver from within Harvester, then link Harvester and Rancher. Then hop over to the Rancher UI and create the second RKE2 cluster with the Harvester node driver.