In a service definition, the 'portalIP' value has been renamed to 'clusterIP'. 

The problem is that existing configuration files which use this 'portalIP' value will fail when updating to OpenShift 3.4. This is a silent failure where the configuration file will be accepted, it will just not have its the value applied properly.

This means existing components which rely on the behaviour of 'portalIP' will start to fail.

We should continue to handle 'portalIP' values in a backwards compatible manner.

This will fail to properly set the portalIP to 'None':

- apiVersion: v1
  kind: Service
  metadata:
    labels:
      metrics-infra: hawkular-cassandra
      name: hawkular-cassandra
    name: hawkular-cassandra-nodes
  spec:
    portalIP: None
    ports:
    - name: cql-port
      port: 9042
    - name: thift-port
      port: 9160
    - name: tcp-port
      port: 7000
    - name: ssl-port
      port: 7001
    selector:
      type: hawkular-cassandra

While this will properly set the clusterIP to 'None':
- apiVersion: v1
  kind: Service
  metadata:
    labels:
      metrics-infra: hawkular-cassandra
      name: hawkular-cassandra
    name: hawkular-cassandra-nodes
  spec:
    clusterIP: None
    ports:
    - name: cql-port
      port: 9042
    - name: thift-port
      port: 9160
    - name: tcp-port
      port: 7000
    - name: ssl-port
      port: 7001
    selector:
      type: hawkular-cassandra