Bug 1449858

Description of problem:

I am using the same inventory and playbood to deploy logging and just changing the logging version.   

3.5.0 installs and operates great.  

3.6.0 - the kibana pod goes into CrashLoopBackoff with the kibana-proxy container never starting.   The container logs only show this:

Could not read TLS opts from secret/server-tls.json


Version-Release number of selected component (if applicable): 3.6.0


How reproducible: Always


Steps to Reproduce:
1. Install OCP 3.6.65
2. Run ansible playbook openshift-ansible/playbooks/byo/openshift-cluster/openshift-logging.yml with the following inventory

openshift_logging_install_logging=true
openshift_logging_use_ops=false
openshift_logging_master_url=https://ec2-54-202-103-83.us-west-2.compute.amazonaws.com:8443
openshift_logging_master_public_url=https://ec2-54-202-103-83.us-west-2.compute.amazonaws.com:8443
openshift_logging_kibana_hostname=kibana.0510-o0j.qe.rhcloud.com
openshift_logging_namespace=logging
openshift_logging_image_prefix=registry.ops.openshift.com/openshift3/
openshift_logging_image_version=3.6.0
openshift_logging_es_pvc_dynamic=true
openshift_logging_es_pvc_size=25Gi


Actual results:

If openshift_logging_image_version is 3.6.0 the kibana-proxy container will not start with the error above.   If the version is 3.5.0, all is good.

Expected results:

Logging deploys and all pods started.