Add prometheus metrics

charts/rancher-backup/templates/deployment.yaml

@@ -19,6 +19,10 @@ spec:
       annotations:
         checksum/s3: {{ include (print $.Template.BasePath "/s3-secret.yaml") . | sha256sum }}
         checksum/pvc: {{ include (print $.Template.BasePath "/pvc.yaml") . | sha256sum }}
+        {{- if .Values.monitoring.metrics.enabled }}
+        prometheus.io/port: "metrics"
+        prometheus.io/scrape: "true"
+        {{ end }}
     spec:
       serviceAccountName: {{ include "backupRestore.serviceAccountName" . }}
       {{- if .Values.imagePullSecrets }}
@@ -32,6 +36,8 @@ spec:
       - name: {{ .Chart.Name }}
         image: {{ template "system_default_registry" . }}{{ .Values.image.repository }}:{{ .Values.image.tag }}
         imagePullPolicy: {{ default "Always" .Values.imagePullPolicy }}
+        ports:
+        - containerPort: 8080
         args:
 {{- if .Values.debug }}
         - "--debug"
@@ -54,6 +60,10 @@ spec:
         - name: NO_PROXY
           value: {{ .Values.noProxy }}
           {{- end }}
+          {{- if .Values.monitoring.metrics.enabled }}
+        - name: METRICS_SERVER
+          value: "true"
+          {{ end }}
           {{- if .Values.persistence.enabled }}
         - name: DEFAULT_PERSISTENCE_ENABLED
           value: "persistence-enabled"

charts/rancher-backup/templates/hardened.yaml

@@ -117,6 +117,8 @@ metadata:
   namespace: {{ .Release.Namespace }}
 spec:
   podSelector: {}
+  ingress:
+    - {}
   egress:
     - {}
   policyTypes:

charts/rancher-backup/templates/service-monitor.yaml

@@ -0,0 +1,30 @@
+{{ if and (.Capabilities.APIVersions.Has "monitoring.coreos.com/v1") .Values.monitoring.serviceMonitor.enabled }}
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: {{ include "backupRestore.fullname" . }}
+  namespace: {{ .Release.Namespace }}
+  labels:
+{{- include "backupRestore.labels" . | nindent 4 }}
+{{- with .Values.monitoring.serviceMonitor.additionalLabels }}
+    {{- toYaml . | nindent 4 }}
+{{- end }}
+spec:
+  selector:
+    matchLabels:
+{{- include "backupRestore.labels" . | nindent 6 }}
+  endpoints:
+  - port: http
+    path: /metrics
+    {{- with .Values.monitoring.serviceMonitor.metricRelabelings }}
+    metricRelabelings:
+      {{- toYaml . | nindent 6 }}
+    {{- end }}
+    {{- with .Values.monitoring.serviceMonitor.relabelings }}
+    relabelings:
+      {{- toYaml . | nindent 4 }}
+    {{- end }}
+  namespaceSelector:
+    matchNames:
+    - {{ .Release.Namespace }}
+{{- end }}

charts/rancher-backup/templates/service.yaml

@@ -0,0 +1,24 @@
+{{ if .Values.monitoring.metrics.enabled }}
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "backupRestore.fullname" . }}
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "backupRestore.labels" . | nindent 4 }}
+  annotations:
+    prometheus.io/path: /metrics
+    prometheus.io/port: /8080
+    prometheus.io/scrape: "true"
+spec:
+  type: ClusterIP
+  clusterIP: None
+  ports:
+    - port: 8080
+      targetPort: 8080
+      protocol: TCP
+      name: http
+  selector:
+    app.kubernetes.io/name: {{ .Release.Name }}
+    app.kubernetes.io/instance: {{ .Release.Name }}
+{{ end }}

charts/rancher-backup/values.yaml

@@ -79,3 +79,14 @@ imagePullPolicy: "Always"
 ## Optional array of imagePullSecrets containing private registry credentials
 ## Ref: https://kubernetes.io/docs/tasks/configure-pod-container/pull-image-private-registry/
 imagePullSecrets: []
+
+monitoring:
+  metrics:
+    enabled: false
+
+  serviceMonitor:
+    enabled: false
+
+    additionalLabels: {}
+    metricRelabelings: []
+    relabelings: []

cmd/operator/main.go

@@ -21,6 +21,7 @@ var (
 	LocalBackupStorageLocation      = "/var/lib/backups" // local within the pod, this is the mountPath for PVC
 	KubeConfig                      string
 	OperatorPVEnabled               string
+	MetricsServerEnabled            string
 	OperatorS3BackupStorageLocation string
 	ChartNamespace                  string
 	Debug                           bool
@@ -36,6 +37,7 @@ func init() {
 	OperatorPVEnabled = os.Getenv("DEFAULT_PERSISTENCE_ENABLED")
 	OperatorS3BackupStorageLocation = os.Getenv("DEFAULT_S3_BACKUP_STORAGE_LOCATION")
 	ChartNamespace = os.Getenv("CHART_NAMESPACE")
+	MetricsServerEnabled = os.Getenv("METRICS_SERVER")
 }
 
 func main() {
@@ -60,6 +62,9 @@ func main() {
 	backuputil.SetDevMode(dm != "")
 	runOptions := operator.RunOptions{
 		OperatorPVCEnabled:              OperatorPVEnabled != "",
+		MetricsServerEnabled:            MetricsServerEnabled != "",
+		MetricsPort:                     8080,
+		MetricsIntervalSeconds:          60,
 		OperatorS3BackupStorageLocation: OperatorS3BackupStorageLocation,
 		ChartNamespace:                  ChartNamespace,
 		LocalDriverPath:                 "",