Red Hat OpenShift Container Platform master scaleup problem with 3.4 playbooks and existing old 3.x master

Issue

• We are trying to scale-up a new master but are having the following issue with the playbook. At some point it tries to delete generated certificates with a with-items in ansible. But this step does not check if the files were generated on the first hand. It seems like our "old" first master does not have the following files, thus does not generate the files for the new master. Then the deletion step comes along and crashes:

failed: [new-master.example.com -> exisiting-master.example.com] (item=service-signer.crt) => {
    "failed": true, 
    "item": "service-signer.crt", 
    "path": "/etc/origin/generated-configs/new-master.example.com/service-signer.crt", 
    "state": "absent"
}