[ceph-osd] Get ceph crush rack names from node lables

This is to enhance the ceph rack creation logic to read the rack names
from node labels so that could avoid providing rack names as chart overrides.

Change-Id: I5a29584d105fba068516d396ada90d00e2aab49c

ceph-osd/templates/bin/osd/ceph-volume/_common.sh.tpl

@@ -25,7 +25,16 @@ export PS4='+${BASH_SOURCE:+$(basename ${BASH_SOURCE}):${LINENO}:}${FUNCNAME:+${
 : "${OSD_JOURNAL_SIZE:=$(awk '/^osd_journal_size/{print $3}' ${CEPH_CONF}.template)}"
 : "${OSD_WEIGHT:=1.0}"
 
+eval CRUSH_FAILURE_DOMAIN_NAME_FROM_NODE_LABEL=$(kubectl get node  ${HOSTNAME} -o json| jq -r '.metadata.labels.rack')
 eval CRUSH_FAILURE_DOMAIN_TYPE=$(cat /etc/ceph/storage.json | python -c 'import sys, json; data = json.load(sys.stdin); print(json.dumps(data["failure_domain"]))')
+
+if [ ${CRUSH_FAILURE_DOMAIN_NAME_FROM_NODE_LABEL} == "null" ]; then
+
+  eval CRUSH_FAILURE_DOMAIN_NAME=$(cat /etc/ceph/storage.json | python -c 'import sys, json; data = json.load(sys.stdin); print(json.dumps(data["failure_domain_name"]))')
+else
+  CRUSH_FAILURE_DOMAIN_NAME=${CRUSH_FAILURE_DOMAIN_NAME_FROM_NODE_LABEL}
+fi
+
 eval CRUSH_FAILURE_DOMAIN_NAME=$(cat /etc/ceph/storage.json | python -c 'import sys, json; data = json.load(sys.stdin); print(json.dumps(data["failure_domain_name"]))')
 eval CRUSH_FAILURE_DOMAIN_BY_HOSTNAME=$(cat /etc/ceph/storage.json | python -c 'import sys, json; data = json.load(sys.stdin); print(json.dumps(data["failure_domain_by_hostname"]))')
 eval CRUSH_FAILURE_DOMAIN_FROM_HOSTNAME_MAP=$(cat /etc/ceph/storage.json | jq '.failure_domain_by_hostname_map."'$HOSTNAME'"')

ceph-osd/templates/daemonset-osd.yaml

@@ -12,6 +12,40 @@ See the License for the specific language governing permissions and
 limitations under the License.
 */}}
 
+{{- if .Values.manifests.daemonset_osd }}
+{{- $envAll := . }}
+
+{{- $serviceAccountName := (printf "%s" .Release.Name) }}
+{{ tuple . "osd" $serviceAccountName | include "helm-toolkit.snippets.kubernetes_pod_rbac_serviceaccount" }}
+
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: {{ $serviceAccountName }}
+rules:
+  - apiGroups:
+      - ""
+    resources:
+      - nodes
+    verbs:
+      - get
+      - list
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: {{ $serviceAccountName }}
+subjects:
+  - kind: ServiceAccount
+    name: {{ $serviceAccountName }}
+    namespace: {{ .Release.Namespace }}
+roleRef:
+  kind: ClusterRole
+  name: {{ $serviceAccountName }}
+  apiGroup: rbac.authorization.k8s.io
+{{- end }}
+
 {{- define "ceph.osd.daemonset" }}
 {{- $daemonset := index . 0 }}
 {{- $configMapName := index . 1 }}
@@ -460,7 +494,6 @@ spec:
 {{- $daemonset := .Values.daemonset.prefix_name }}
 {{- $configMapName := (printf "%s-%s" .Release.Name "etc") }}
 {{- $serviceAccountName := (printf "%s" .Release.Name) }}
-{{ tuple . "osd" $serviceAccountName | include "helm-toolkit.snippets.kubernetes_pod_rbac_serviceaccount" }}
 {{- $daemonset_yaml := list $daemonset $configMapName $serviceAccountName . | include "ceph.osd.daemonset" | toString | fromYaml }}
 {{- $configmap_yaml := "ceph.osd.configmap.etc" }}
 {{- list $daemonset $daemonset_yaml $configmap_yaml $configMapName . | include "ceph.utils.osd_daemonset_overrides" }}

ceph-osd/values.yaml

@@ -209,6 +209,8 @@ conf:
     # rack_replicated_rule you would specify "rack" as the `failure_domain` to use.
     # `failure_domain`: Set the CRUSH bucket type for your OSD to reside in. See the supported CRUSH configuration
     #  as listed here: Supported CRUSH configuration is listed here: http://docs.ceph.com/docs/nautilus/rados/operations/crush-map/
+    #  if failure domain is rack then it will check for node label "rack" and get the value from it to create the rack, if there
+    #  is no label rack then it will use following options.
     # `failure_domain_by_hostname`: Specify the portion of the hostname to use for your failure domain bucket name.
     # `failure_domain_by_hostname_map`: Explicit mapping of hostname to failure domain, as a simpler alternative to overrides.
     # `failure_domain_name`: Manually name the failure domain bucket name. This configuration option should only be used