Mismatch fqdn vs hostname.
Ensure output of 'hostname' != 'hostname -f' on each host.
EG:
$ hostname
myhost
$ hostname -f
myhost.example.com
Install 3.9 as normal.
Run upgrade, should fail early with hostname mismatch.
Set each host to have output of 'hostname' == 'hostname -f' via hostnamectl.
Rerun upgrade, upgrade should complete without issue.
Mismatch hostname vs openshift_hostname
Ensure openshift_hostname is set in inventory for each host, and that it is resolvable. Also ensure that openshift_hostname != 'hostname' output. This might require custom DNS or on aws (without using cloudprovider) you might be able to set openshift_hostname to external ec2 names (depending on if each host can connect to each other via external names).
Should fail same as scenario 1.
Set openshift_kubelet_name_override to value openshift_hostname.
Remove openshift_hostname from inventory for each host.
Run upgrade, it should complete successfully.
Verify on each node the last line of node-config.yml contains "nodeName: $openshift_kubelet_name_override"
Verify you can scale up a node to cluster from scenario 2.
Attempt to set openshift_kubelet_name_override for the scale up host.
Install should fail.
Remove openshift_kubelet_name_override only from new host, run scaleup.
Install should succeed.
On both 3.10 and 3.11 branches, try to install a new cluster with openshift_kubelet_name_override set. Installs should fail. Installs should proceed without this value.