rke2-capd workload example has no capd image_name, capd tries to pull kindest/node:<rke2 version>
Summary
environment-values/workload-clusters/rke2-capd doesn't set cluster.capd.image_name, so capd builds a kindest/node image name from the rke2 version, which doesn't exist. The workload machine container never gets created. Seen on 1.7.7 and 1.8.0. Related to #4844
Steps to reproduce
- Deploy a management cluster from
environment-values/rke2-capd - Copy
environment-values/workload-clusters/rke2-capd, set cluster_virtual_ip - Run
apply-workload-cluster.shwith it
What is the current bug behavior?
Only the lb container comes up. The machine status just says waiting for bootstrap data, the actual error is only in the capd controller logs.
What is the expected correct behavior?
Workload machine uses the rke2-in-docker image like the management cluster.
Relevant logs and/or screenshots
capd-controller-manager:
"Reconciler error" err="failed to create worker DockerMachine: error pulling container image kindest/node:v1.36.5_rke2r1: failure pulling container image: Error response from daemon: failed to resolve reference \"docker.io/kindest/node:v1.36.5_rke2r1\"Possible fixes
The mgmt example sets the image (environment-values/rke2-capd/values.yaml:23), the workload example doesn't.
Setting cluster.capd.image_name in the workload example to the same image fixes it for us (the already created machine has to be deleted once). We'll send an MR.