Description of problem:

After installing cluster-logging through the marketplace, creation of clusterlogging CRDs results in a validation error in the CLO logs:

time="2019-01-29T20:51:53Z" level=error msg="error syncing key (openshift-logging/storageclassname): Unable to create or update logstore: Failure creating Elasticsearch CR: Elasticsearch.logging.openshift.io \"elasticsearch\" is invalid: []: Invalid value: map[string]interface {}{\"metadata\":map[string]interface {}{\"name\":\"elasticsearch\", \"namespace\":\"openshift-logging\", \"creationTimestamp\":\"2019-01-29T20:51:53Z\", \"ownerReferences\":[]interface {}{map[string]interface {}{\"apiVersion\":\"logging.openshift.io/v1alpha1\", \"kind\":\"ClusterLogging\", \"name\":\"storageclassname\", \"uid\":\"9994e1d5-2407-11e9-a060-06d58bf97baa\", \"controller\":true}}, \"generation\":1, \"uid\":\"b53ef32e-2407-11e9-a060-06d58bf97baa\"}, \"spec\":map[string]interface {}{\"managementState\":\"Managed\", \"nodeSpec\":map[string]interface {}{\"image\":\"quay.io/openshift/origin-logging-elasticsearch5:latest\", \"resources\":map[string]interface {}{}}, \"nodes\":[]interface {}{map[string]interface {}{\"nodeSpec\":map[string]interface {}{\"resources\":map[string]interface {}{}}, \"roles\":[]interface {}{\"client\", \"data\", \"master\"}, \"storage\":map[string]interface {}{\"size\":\"20G\"}, \"nodeCount\":3}}, \"redundancyPolicy\":\"SingleRedundancy\"}, \"status\":map[string]interface {}{\"shardAllocationEnabled\":\"\", \"clusterHealth\":\"\", \"conditions\":interface {}(nil), \"nodes\":interface {}(nil), \"pods\":interface {}(nil)}, \"apiVersion\":\"logging.openshift.io/v1alpha1\", \"kind\":\"Elasticsearch\"}: validation failure list:\nspec.nodes.roles in body must be of type object: \"array\""


Version-Release number of selected component (if applicable): OCP is 4.0.0-0.nightly-2019-01-29-025207.   Logging is latest via marketplace as of Jan 29


How reproducible: Always


Steps to Reproduce:
1. Install the logging operators from the marketplace
2. Create a clusterlogging resource in openshift-logging (hack/cr.yaml will have the issue)
3. oc logs on the cluster-logging-operator pod

Actual results:

Error as above