Units living in management cluster use workload cluster proxy config

Summary

In case management and workload clusters have different proxy configuration (for example when deployed in different environments), some units created for workload cluster deployment may fail because they are using workload cluster proxy even if they actually are run into the management cluster.

Details

I fell into this issue when upgrading from Sylva 1.3.10 to 1.3.14 in baremetal environment.

• os-images-info was not triggered in 1.3.10 but is created in 1.3.14 in case of capm3 deployment.
• os-image-info kube-job pod runs with workload cluster proxy config but fails because it runs into management cluster.