Hello! I am using the examples to deploy a child RKE2 cluster via TFR2P and Harvester as the infrastructure and cloud provider. I faced the following issue: the cloud provider config is passed as follows:

machine_selector_config {
  config = jsonencode({
    cloud-provider-config: local_file.harvester-kube-config.content
    cloud-provider-name: "harvester"
  })
}

however, I need to manually patch each cluster member using the following, because the harvester-cloud-provider pod was not able to find any of the cluster nodes:

kubectl patch node tofu-test-allroles-fgqkm-xxxxx \
  -p '{"spec":{"providerID":"<harvester://harvester-public/tofu-test-allroles-fgqkm-xxxxx>"}}'

as well as I had to remove the taint from the master node, otherwise calico would not start. the taint being:

<http://node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule|node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule>

which is because the master node of the cluster hasn't initialized properly because the of the CCM not being able to find it --> hence the

kubectl patch

above. As per the example in the docs, I am using

fleet_namespace = "fleet-default"

and

harvester_config.vm_namespace = "harvester-public"

in machine_config_v2. Any idea why this happens? Anyone encountered this strange behavior in the past? All feedback is welcome 🙂 P.s. if this is not the correct channel to post this, please advise. Ronald