Description of problem: VM cluster inside of OpenStack has bind running to provide DNS for custom cluster domain. Hostnames and DNS configuration is set correctly on all VM hosts in cluster and resolves backwards and forwards. However during installation the nodes do not get the correct hostname from the host, instead they get the IP address. The local instance hostname on the node itself does not change. The OpenShift install is being run *without* the OpenStack provider. Version-Release number of selected component (if applicable): git master openshift-ansible How reproducible: 99% of the time, however very rarely it will either take the correct hostname or even less often take the original hostname of the actual VM instance. Steps to Reproduce: 1. Have working OpenStack cluster, deploy n VM instances. 2. Install and configure bind within the private network of the instances, ensure PEERDNS is set to false to ensure we take the local DNS settings. 3. Install OpenShift via advanced installer, listing the nodes using their correct hostnames, with not other overrides or hostname specific parameters. Actual results: OpenShift nodes (as shown in `oc get no`) will have their hostnames set to their IP addresses. Expected results: OpenShift nodes are set to their correct hostnames as listed in the inventory.
I think it should be possible to avoid this by setting something like: [nodes] master.example.com openshift_hostname=master.example.com ... But indeed the playbooks should be smart enough not to require this.