This is a clone of issue
OCPBUGS-6018. The following is the description of the original issue:
This is a public clone of OCPBUGS-3821
The MCO can sometimes render a rendered-config in the middle of an upgrade with old MCs, e.g.:
- the containerruntimeconfigcontroller creates a new containerruntimeconfig due to the update
- the template controller finishes re-creating the base configs
- the kubeletconfig errors long enough and doesn't finish until after 2
This will cause the render controller to create a new rendered MC that uses the OLD kubeletconfig-MC, which at best is a double reboot for 1 node, and at worst block the update and break maxUnavailable nodes per pool.