Description of problem:

The nodeSelector are changed in the logging-es-ops deploymentconfig after upgrade. that logging-es pod couldn't be started.

1) nodeSelector before Upgrade:
cat elasticsearch-dc-before-upgrade.json | jq '.items[].metadata.name, .items[].spec.template.spec.nodeSelector'
"logging-es-data-master-ajbqhp8h"
"logging-es-data-master-telafmeq"
"logging-es-ops-data-master-0fr84k1a"
"logging-es-ops-data-master-9961o92h"
"logging-es-ops-data-master-o7nhcbo4"
{
  "logging-es-node": "1"
}
{
  "logging-es-node": "0"
}
{
  "logging-es-ops-node": "2"
}
{
  "logging-es-ops-node": "0"
}
{
  "logging-es-ops-node": "1"
}


2) Logging Inventory used for upgrade
openshift_logging_install_logging=true
openshift_logging_es_cluster_size=2
openshift_logging_es_number_of_replicas=1
openshift_logging_es_number_of_shards=1
openshift_logging_es_memory_limit=2Gi
openshift_logging_es_nodeselector={"node-role.kubernetes.io/infra": "true"}

openshift_logging_use_ops=true
openshift_logging_es_ops_cluster_size=3
openshift_logging_es_ops_number_of_replicas=1
openshift_logging_es_ops_number_of_shards=1
openshift_logging_es_ops_memory_limit=2Gi
openshift_logging_es_ops_nodeselector={"node-role.kubernetes.io/compute": "true"}

openshift_logging_elasticsearch_storage_type=hostmount

3)nodeSelector after Upgrade
cat elasticsearch-dc-after.json | jq '.items[].metadata.name, .items[].spec.template.spec.nodeSelector'
"logging-es-data-master-ajbqhp8h"
"logging-es-data-master-telafmeq"
"logging-es-ops-data-master-0fr84k1a"
"logging-es-ops-data-master-9961o92h"
"logging-es-ops-data-master-o7nhcbo4"
{
  "logging-es-node": "1"
}
{
  "logging-es-node": "0"
}
{
  "logging-es-node": "2"
}
{
  "logging-es-node": "0"
}
{
  "logging-es-node": "1"


Version-Release number of selected component (if applicable):
openshift3/ose-ansible:v3.11.135 

How reproducible:
Always

Steps to Reproduce:
1. deploy logging using openshift_logging_use_ops=true

openshift_logging_install_logging=true
openshift_logging_es_cluster_size=2
openshift_logging_es_number_of_replicas=1
openshift_logging_es_number_of_shards=1
openshift_logging_es_memory_limit=2Gi
openshift_logging_es_nodeselector={"node-role.kubernetes.io/infra": "true"}

openshift_logging_use_ops=true
openshift_logging_es_ops_cluster_size=3
openshift_logging_es_ops_number_of_replicas=1
openshift_logging_es_ops_number_of_shards=1
openshift_logging_es_ops_memory_limit=2Gi
openshift_logging_es_ops_nodeselector={"node-role.kubernetes.io/compute": "true"}

2. Add hostpath volume and nodeSelector to ES and ES-Ops deployment configure
"logging-es-data-master-ajbqhp8h"
"logging-es-data-master-telafmeq"
"logging-es-ops-data-master-0fr84k1a"
"logging-es-ops-data-master-9961o92h"
"logging-es-ops-data-master-o7nhcbo4"
{
  "logging-es-node": "1"
}
{
  "logging-es-node": "0"
}
{
  "logging-es-ops-node": "2"
}
{
  "logging-es-ops-node": "0"
}
{
  "logging-es-ops-node": "1"
}

3. Upgrade to latest version using openshift3/ose-ansible:v3.11.135 

Actual results:
The logging-es-ops pod couldn't be started, as the nodeSelector are changed after upgrade
cat elasticsearch-dc-after.json | jq '.items[].metadata.name, .items[].spec.template.spec.nodeSelector'
"logging-es-data-master-ajbqhp8h"
"logging-es-data-master-telafmeq"
"logging-es-ops-data-master-0fr84k1a"
"logging-es-ops-data-master-9961o92h"
"logging-es-ops-data-master-o7nhcbo4"
{
  "logging-es-node": "1"
}
{
  "logging-es-node": "0"
}
{
  "logging-es-node": "2"
}
{
  "logging-es-node": "0"
}
{
  "logging-es-node": "1"