Create kepler scrape config and dashboard only when power monitoring …

…is enabled

api/v1beta1/telemetry_consts.go

@@ -32,6 +32,8 @@ const (
 	PauseBetweenWatchAttempts = time.Duration(60) * time.Second
 	// DefaultKeplerPort -
 	DefaultKeplerPort = 8888
+	// Dataplane power monitoring service name
+	TelemetryPowerMonitoring = "telemetry-power-monitoring"
 )
 
 // PrometheusReplicas -

controllers/metricstorage_controller.go

@@ -439,7 +439,7 @@ func (r *MetricStorageReconciler) reconcileNormal(
 		instance.Status.Conditions.MarkTrue(telemetryv1.DashboardDefinitionReadyCondition, telemetryv1.DashboardsNotEnabledMessage)
 		instance.Status.Conditions.MarkTrue(telemetryv1.DashboardPluginReadyCondition, telemetryv1.DashboardsNotEnabledMessage)
 	} else {
-		if res, err := r.createDashboardObjects(ctx, instance, eventHandler); err != nil {
+		if res, err := r.createDashboardObjects(ctx, instance, helper, eventHandler); err != nil {
 			return res, err
 		}
 	}
@@ -601,7 +601,7 @@ func (r *MetricStorageReconciler) createScrapeConfigs(
 		return ctrl.Result{}, err
 	}
 
-	connectionInfo, err := getComputeNodesConnectionInfo(instance, helper)
+	connectionInfo, err := getComputeNodesConnectionInfo(instance, helper, telemetry.ServiceName)
 	if err != nil {
 		Log.Info(fmt.Sprintf("Cannot get compute node connection info. Scrape configs not created. Error: %s", err))
 	}
@@ -624,18 +624,26 @@ func (r *MetricStorageReconciler) createScrapeConfigs(
 		return ctrl.Result{}, err
 	}
 
+	connectionInfo, err = getComputeNodesConnectionInfo(instance, helper, telemetryv1.TelemetryPowerMonitoring)
+	if err != nil {
+		Log.Info(fmt.Sprintf("Cannot get compute node connection info. Scrape configs not created. Error: %s", err))
+	}
+
 	// kepler scrape endpoints
 	keplerEndpoints, _ := getKeplerTargets(connectionInfo)
 	if err != nil {
 		Log.Info(fmt.Sprintf("Cannot get Kepler targets. Scrape configs not created. Error: %s", err))
 	}
 
-	// Kepler ScrapeConfig for non-tls nodes
-	keplerServiceName := fmt.Sprintf("%s-kepler", telemetry.ServiceName)
-	err = r.createServiceScrapeConfig(ctx, instance, Log, "Kepler",
-		keplerServiceName, keplerEndpoints, false) // Currently Kepler doesn't support TLS so tlsEnabled is set to false
-	if err != nil {
-		return ctrl.Result{}, err
+	// keplerEndpoint is reported as empty slice when telemetry-power-monitoring service is not enabled
+	if len(keplerEndpoints) > 0 {
+		// Kepler ScrapeConfig for non-tls nodes
+		keplerServiceName := fmt.Sprintf("%s-kepler", telemetry.ServiceName)
+		err = r.createServiceScrapeConfig(ctx, instance, Log, "Kepler",
+			keplerServiceName, keplerEndpoints, false) // Currently Kepler doesn't support TLS so tlsEnabled is set to false
+		if err != nil {
+			return ctrl.Result{}, err
+		}
 	}
 
 	instance.Status.Conditions.MarkTrue(telemetryv1.ScrapeConfigReadyCondition, condition.ReadyMessage)
@@ -673,7 +681,7 @@ func getKeplerTargets(nodes []ConnectionInfo) ([]string, []string) {
 	return tls, nonTLS
 }
 
-func (r *MetricStorageReconciler) createDashboardObjects(ctx context.Context, instance *telemetryv1.MetricStorage, eventHandler handler.EventHandler) (ctrl.Result, error) {
+func (r *MetricStorageReconciler) createDashboardObjects(ctx context.Context, instance *telemetryv1.MetricStorage, helper *helper.Helper, eventHandler handler.EventHandler) (ctrl.Result, error) {
 	Log := r.GetLogger(ctx)
 	uiPluginObj := &obsui.UIPlugin{
 		ObjectMeta: metav1.ObjectMeta{
@@ -766,7 +774,14 @@ func (r *MetricStorageReconciler) createDashboardObjects(ctx context.Context, in
 			"grafana-dashboard-openstack-node":     dashboards.OpenstackNode(datasourceName),
 			"grafana-dashboard-openstack-vm":       dashboards.OpenstackVM(datasourceName),
 			"grafana-dashboard-openstack-rabbitmq": dashboards.OpenstackRabbitmq(datasourceName),
-			"grafana-dashboard-openstack-kepler":   dashboards.OpenstackKepler(datasourceName),
+		}
+
+		// atleast one nodeset must have "telemetry-power-monitoring" service enabled for kepler dashboard to be created
+		connectionInfo, err := getComputeNodesConnectionInfo(instance, helper, telemetryv1.TelemetryPowerMonitoring)
+		if err != nil {
+			Log.Info(fmt.Sprintf("Cannot get compute node connection info. Power monitoring dashboard not created. Error: %s", err))
+		} else if len(connectionInfo) > 0 {
+			dashboardCMs["grafana-dashboard-openstack-kepler"] = dashboards.OpenstackKepler(datasourceName)
 		}
 
 		for dashboardName, desiredCM := range dashboardCMs {
@@ -838,6 +853,7 @@ func (r *MetricStorageReconciler) ensureWatches(
 func getComputeNodesConnectionInfo(
 	instance *telemetryv1.MetricStorage,
 	helper *helper.Helper,
+	telemetryServiceName string,
 ) ([]ConnectionInfo, error) {
 	ipSetList, err := getIPSetList(instance, helper)
 	if err != nil {
@@ -855,24 +871,23 @@ func getComputeNodesConnectionInfo(
 			return []ConnectionInfo{}, err
 		}
 		nodeSetGroup := inventory.Groups[secret.Labels["openstackdataplanenodeset"]]
-		containsTelemetry := false
+		containsTargetService := false
 		for _, svc := range nodeSetGroup.Vars["edpm_services"].([]interface{}) {
-			if svc.(string) == "telemetry" {
-				containsTelemetry = true
+			if svc.(string) == telemetryServiceName {
+				containsTargetService = true
 			}
 		}
-		if !containsTelemetry {
-			// Telemetry isn't deployed on this nodeset
-			// there is no reason to include these nodes
-			// for scraping by prometheus
+		if !containsTargetService {
+			// If Telemetry|TelemetryPowerMonitoring isn't
+			// deployed on this nodeset there is no reason
+			// to include these nodes for scraping by prometheus
 			continue
 		}
 		for name, item := range nodeSetGroup.Hosts {
 			namespacedName := &types.NamespacedName{
 				Name:      name,
 				Namespace: instance.GetNamespace(),
 			}
-
 			if len(ipSetList.Items) > 0 {
 				// if we have IPSets, lets go to search for the IPs there
 				address, _ = getAddressFromIPSet(instance, &item, namespacedName, helper)

kuttl-test.yaml

@@ -28,3 +28,17 @@ testDirs:
   - tests/kuttl/suites/tls/
 suppress:
   - events                     # Remove spammy event logs
+commands:
+  - script: |
+      if [ ! -f ansibleee-ssh-key-id_rsa ]; then
+          ssh-keygen -f ansibleee-ssh-key-id_rsa -N "" -t rsa -b 4096
+      fi
+      oc create secret generic dataplane-ansible-ssh-private-key-secret \
+      --save-config \
+      --dry-run=client \
+      --from-file=authorized_keys=ansibleee-ssh-key-id_rsa.pub \
+      --from-file=ssh-privatekey=ansibleee-ssh-key-id_rsa \
+      --from-file=ssh-publickey=ansibleee-ssh-key-id_rsa.pub \
+      -n telemetry-kuttl-tests \
+      -o yaml | \
+      oc apply -f -

tests/kuttl/suites/metricstorage/deps/dnsmasq.yaml

@@ -0,0 +1,25 @@
+apiVersion: network.openstack.org/v1beta1
+kind: DNSMasq
+metadata:
+  name: dnsmasq
+  namespace: telemetry-kuttl-tests
+spec:
+  replicas: 1
+  options:
+  - key: server
+    values:
+    - 192.168.122.1
+  - key: local
+    values:
+    - '/example.com/'
+  debug:
+    service: false
+  override:
+    service:
+      metadata:
+        annotations:
+          metallb.universe.tf/address-pool: ctlplane
+          metallb.universe.tf/allow-shared-ip: ctlplane
+          metallb.universe.tf/loadBalancerIPs: 192.168.122.80
+      spec:
+        type: ClusterIP

tests/kuttl/suites/metricstorage/deps/inventory

@@ -0,0 +1,119 @@
+openstack-edpm-ipam:
+    vars:
+        edpm_network_config_template: |
+            ---
+            {% set mtu_list = [ctlplane_mtu] %}
+            {% for network in nodeset_networks %}
+            {{ mtu_list.append(lookup('vars', networks_lower[network] ~ '_mtu')) }}
+            {%- endfor %}
+            {% set min_viable_mtu = mtu_list | max %}
+            network_config:
+            - type: interface
+              name: nic1
+              use_dhcp: false
+            - type: interface
+              name: nic2
+              use_dhcp: false
+              addresses:
+              - ip_netmask: {{ ctlplane_ip }}/{{ ctlplane_cidr }}
+              routes:
+              - default: true
+                next_hop: {{ ctlplane_gateway_ip }}
+
+            - type: linux_bond
+              name: bond_api
+              use_dhcp: false
+              bonding_options: "mode=active-backup"
+              dns_servers: {{ ctlplane_dns_nameservers }}
+              members:
+              - type: interface
+                name: nic3
+
+            - type: vlan
+              vlan_id: {{ lookup('vars', networks_lower['internalapi'] ~ '_vlan_id') }}
+              device: bond_api
+              addresses:
+              - ip_netmask: {{ lookup('vars', networks_lower['internalapi'] ~ '_ip') }}/{{ lookup('vars', networks_lower['internalapi'] ~ '_cidr') }}
+
+            - type: vlan
+              vlan_id: {{ lookup('vars', networks_lower['storage'] ~ '_vlan_id') }}
+              device: bond_api
+              addresses:
+              - ip_netmask: {{ lookup('vars', networks_lower['storage'] ~ '_ip') }}/{{ lookup('vars', networks_lower['storage'] ~ '_cidr') }}
+        edpm_service_types:
+            - bootstrap
+            - download-cache
+            - reboot-os
+            - configure-ovs-dpdk
+            - configure-network
+            - validate-network
+            - install-os
+            - configure-os
+            - ssh-known-hosts
+            - run-os
+            - install-certs
+            - ovn
+            - neutron-ovn
+            - neutron-metadata
+            - neutron-sriov
+            - libvirt
+            - nova
+            - telemetry
+            - telemetry-power-monitoring
+        edpm_services:
+            - bootstrap
+            - download-cache
+            - reboot-os
+            - configure-ovs-dpdk
+            - configure-network
+            - validate-network
+            - install-os
+            - configure-os
+            - ssh-known-hosts
+            - run-os
+            - install-certs
+            - ovn
+            - neutron-ovn-igmp
+            - neutron-metadata
+            - neutron-sriov
+            - libvirt
+            - nova
+            - telemetry
+            - telemetry-power-monitoring
+        edpm_tls_certs_enabled: true
+    hosts:
+        edpm-compute-0:
+            ansible_host: edpm-compute-0
+            canonical_hostname: edpm-compute-0.ctlplane.example.com
+            ctlplane_cidr: 24
+            ctlplane_dns_nameservers:
+                - 192.168.122.80
+            ctlplane_gateway_ip: 192.168.122.1
+            ctlplane_host_routes:
+                - destination: 0.0.0.0/0
+                  nexthop: 192.168.122.1
+            ctlplane_ip: 192.168.122.100
+            ctlplane_mtu: 1500
+            dns_search_domains:
+                - ctlplane.example.com
+                - internalapi.example.com
+                - storage.example.com
+                - tenant.example.com
+            internalapi_cidr: 24
+            internalapi_gateway_ip: null
+            internalapi_host_routes: []
+            internalapi_ip: 172.17.0.100
+            internalapi_mtu: 1496
+            internalapi_vlan_id: 52
+            storage_cidr: 24
+            storage_gateway_ip: null
+            storage_host_routes: []
+            storage_ip: 172.18.0.100
+            storage_mtu: 1496
+            storage_vlan_id: 53
+            tenant_cidr: 24
+            tenant_gateway_ip: null
+            tenant_host_routes: []
+            tenant_ip: 172.19.0.100
+            tenant_mtu: 1496
+            tenant_vlan_id: 54

tests/kuttl/suites/metricstorage/tests/01-assert.yaml

@@ -95,19 +95,6 @@ metadata:
 spec:
   scrapeInterval: 30s
 ---
-apiVersion: monitoring.rhobs/v1alpha1
-kind: ScrapeConfig
-metadata:
-  labels:
-    service: metricStorage
-  name: telemetry-kepler
-  ownerReferences:
-  - kind: MetricStorage
-    name: telemetry-kuttl
-spec:
-  staticConfigs:
-    - {}
----
 apiVersion: observability.openshift.io/v1alpha1
 kind: UIPlugin
 metadata:
@@ -159,9 +146,3 @@ kind: ConfigMap
 metadata:
   name: grafana-dashboard-openstack-rabbitmq
   namespace: openshift-config-managed
----
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: grafana-dashboard-openstack-kepler
-  namespace: openshift-config-managed

tests/kuttl/suites/metricstorage/tests/02-errors.yaml

@@ -36,9 +36,3 @@ kind: ConfigMap
 metadata:
   name: grafana-dashboard-openstack-rabbitmq
   namespace: openshift-config-managed
----
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: grafana-dashboard-openstack-kepler
-  namespace: openshift-config-managed

tests/kuttl/suites/metricstorage/tests/04-assert.yaml

@@ -106,16 +106,3 @@ metadata:
     name: telemetry-kuttl
 spec:
   scrapeInterval: 40s
----
-apiVersion: monitoring.rhobs/v1alpha1
-kind: ScrapeConfig
-metadata:
-  labels:
-    service: metricStorage
-  name: telemetry-kepler
-  ownerReferences:
-  - kind: MetricStorage
-    name: telemetry-kuttl
-spec:
-  staticConfigs:
-    - {}