ACM-14962: init MCOA metrics (#1659)

* update mco api Signed-off-by: Thibault Mange <[email protected]> * add supported and default configs Signed-off-by: Thibault Mange <[email protected]> * add customized variables Signed-off-by: Thibault Mange <[email protected]> * fix test Signed-off-by: Thibault Mange <[email protected]> * add metrics in mcoa enabled Signed-off-by: Thibault Mange <[email protected]> * fix crd name Signed-off-by: Thibault Mange <[email protected]> * update mcoa role Signed-off-by: Thibault Mange <[email protected]> * fix role Signed-off-by: Thibault Mange <[email protected]> * fix role typo Signed-off-by: Thibault Mange <[email protected]> * add write perms for cfg resources Signed-off-by: Thibault Mange <[email protected]> * fix tests Signed-off-by: Thibault Mange <[email protected]> * remove default configs Signed-off-by: Thibault Mange <[email protected]> * dynamic supported configs Signed-off-by: Thibault Mange <[email protected]> * fix Signed-off-by: Thibault Mange <[email protected]> * fix Signed-off-by: Thibault Mange <[email protected]> * fix Signed-off-by: Thibault Mange <[email protected]> * fix supported configs Signed-off-by: Thibault Mange <[email protected]> * revert dynamic supportedConfigs Signed-off-by: Thibault Mange <[email protected]> * fix Signed-off-by: Thibault Mange <[email protected]> * only enabled configs Signed-off-by: Thibault Mange <[email protected]> * change metricsHub var Signed-off-by: Thibault Mange <[email protected]> * change MetricsHubHostname add logic Signed-off-by: Thibault Mange <[email protected]> * fix MetricsHubHostname generation Signed-off-by: Thibault Mange <[email protected]> * simplify mcoa test Signed-off-by: Thibault Mange <[email protected]> --------- Signed-off-by: Thibault Mange <[email protected]>
stolostron · Jan 14, 2025 · 09d061c · 09d061c
1 parent fb915c4
commit 09d061c
Show file tree

Hide file tree

Showing 11 changed files with 553 additions and 23 deletions.
diff --git a/operators/multiclusterobservability/api/v1beta2/multiclusterobservability_types.go b/operators/multiclusterobservability/api/v1beta2/multiclusterobservability_types.go
@@ -68,6 +68,16 @@ type PlatformLogsCollectionSpec struct {
 	Enabled bool `json:"enabled,omitempty"`
 }
 
+// PlatformMetricsCollectionSpec defines the spec for the addon to collect and forward metrics
+// from fleet managed clusters.
+type PlatformMetricsCollectionSpec struct {
+	// Enabled defines a flag to enable/disable the platform metrics collection.
+	//
+	// +optional
+	// +kubebuilder:validation:Optional
+	Enabled bool `json:"enabled,omitempty"`
+}
+
 // PlatformLogsSpec defines the spec for the addon to collect, forward and store logs
 // from fleet managed clusters.
 type PlatformLogsSpec struct {
@@ -79,6 +89,17 @@ type PlatformLogsSpec struct {
 	Collection PlatformLogsCollectionSpec `json:"collection,omitempty"`
 }
 
+// PlatformMetricsSpec defines the spec for the addon to collect, forward and store metrics
+// from fleet managed clusters.
+type PlatformMetricsSpec struct {
+	// Collection defines the spec for the addon to collect and forward metrics
+	// from fleet managed clusters.
+	//
+	// +optional
+	// +kubebuilder:validation:Optional
+	Collection PlatformMetricsCollectionSpec `json:"collection,omitempty"`
+}
+
 // PlatformCapabilitiesSpec defines the observability capabilities managed by the addon
 // for platform components.
 type PlatformCapabilitiesSpec struct {
@@ -88,6 +109,13 @@ type PlatformCapabilitiesSpec struct {
 	// +optional
 	// +kubebuilder:validation:Optional
 	Logs PlatformLogsSpec `json:"logs,omitempty"`
+
+	// Metrics defines the configuration spec for collecting and storing metrics from
+	// platform components running on fleet managed clusters.
+	//
+	// +optional
+	// +kubebuilder:validation:Optional
+	Metrics PlatformMetricsSpec `json:"metrics,omitempty"`
 }
 
 // ClusterLogForwarderSpec defines the spec for the addon to collect and forward logs
@@ -110,6 +138,16 @@ type UserWorkloadLogsCollectionSpec struct {
 	ClusterLogForwarder ClusterLogForwarderSpec `json:"clusterLogForwarder,omitempty"`
 }
 
+// UserWorkloadLogsSpec defines the spec for the addon to collect,forward and store logs
+// from user workloads hosted on fleet managed clusters.
+type UserWorkloadMetricsCollectionSpec struct {
+	// Enabled defines a flag to enable/disable the platform metrics collection.
+	//
+	// +optional
+	// +kubebuilder:validation:Optional
+	Enabled bool `json:"enabled,omitempty"`
+}
+
 // UserWorkloadLogsSpec defines the spec for the addon to collect,forward and store logs
 // from user workloads hosted on fleet managed clusters.
 type UserWorkloadLogsSpec struct {
@@ -121,6 +159,17 @@ type UserWorkloadLogsSpec struct {
 	Collection UserWorkloadLogsCollectionSpec `json:"collection,omitempty"`
 }
 
+// UserWorkloadMetricsSpec defines the spec for the addon to collect, forward and store metrics
+// from user workloads hosted on fleet managed clusters.
+type UserWorkloadMetricsSpec struct {
+	// Collection defines the spec for the addon to collect and forward metrics
+	// from user workloads hosted on fleet managed clusters.
+	//
+	// +optional
+	// +kubebuilder:validation:Optional
+	Collection UserWorkloadMetricsCollectionSpec `json:"collection,omitempty"`
+}
+
 // OpenTelemetryCollectorSpec defines the spec for the addon to collect and forward observability signals
 // using the OpenTelemetryCollector custom resource.
 type OpenTelemetryCollectorSpec struct {
@@ -176,6 +225,14 @@ type UserWorkloadCapabilitiesSpec struct {
 	// +optional
 	// +kubebuilder:validation:Optional
 	Logs UserWorkloadLogsSpec `json:"logs,omitempty"`
+
+	// Metrics defines the spec for the addon to collect, forward and store metrics
+	// from user workloads hosted on fleet managed clusters.
+	//
+	// +optional
+	// +kubebuilder:validation:Optional
+	Metrics UserWorkloadMetricsSpec `json:"metrics,omitempty"`
+
 	// Traces defines the spec for the addon to collect, forward and store traces
 	// from user workloads hosted on fleet managed clusters.
 	//

diff --git a/...rvability/bundle/manifests/multicluster-observability-operator.clusterserviceversion.yaml b/...rvability/bundle/manifests/multicluster-observability-operator.clusterserviceversion.yaml
@@ -49,7 +49,7 @@ metadata:
         }
       ]
     capabilities: Basic Install
-    createdAt: "2024-09-23T12:07:01Z"
+    createdAt: "2024-11-06T16:52:14Z"
     operators.operatorframework.io/builder: operator-sdk-v1.34.2
     operators.operatorframework.io/project_layout: go.kubebuilder.io/v3
   name: multicluster-observability-operator.v0.1.0

diff --git a/...undle/manifests/observability.open-cluster-management.io_multiclusterobservabilities.yaml b/...undle/manifests/observability.open-cluster-management.io_multiclusterobservabilities.yaml
@@ -10036,6 +10036,22 @@ spec:
                                 type: boolean
                             type: object
                         type: object
+                      metrics:
+                        description: |-
+                          Metrics defines the configuration spec for collecting and storing metrics from
+                          platform components running on fleet managed clusters.
+                        properties:
+                          collection:
+                            description: |-
+                              Collection defines the spec for the addon to collect and forward metrics
+                              from fleet managed clusters.
+                            properties:
+                              enabled:
+                                description: Enabled defines a flag to enable/disable
+                                  the platform metrics collection.
+                                type: boolean
+                            type: object
+                        type: object
                     type: object
                   userWorkloads:
                     description: |-
@@ -10066,6 +10082,22 @@ spec:
                                 type: object
                             type: object
                         type: object
+                      metrics:
+                        description: |-
+                          Metrics defines the spec for the addon to collect, forward and store metrics
+                          from user workloads hosted on fleet managed clusters.
+                        properties:
+                          collection:
+                            description: |-
+                              Collection defines the spec for the addon to collect and forward metrics
+                              from user workloads hosted on fleet managed clusters.
+                            properties:
+                              enabled:
+                                description: Enabled defines a flag to enable/disable
+                                  the platform metrics collection.
+                                type: boolean
+                            type: object
+                        type: object
                       traces:
                         description: |-
                           Traces defines the spec for the addon to collect, forward and store traces

diff --git a/...onfig/crd/bases/observability.open-cluster-management.io_multiclusterobservabilities.yaml b/...onfig/crd/bases/observability.open-cluster-management.io_multiclusterobservabilities.yaml
@@ -10020,6 +10020,22 @@ spec:
                                 type: boolean
                             type: object
                         type: object
+                      metrics:
+                        description: |-
+                          Metrics defines the configuration spec for collecting and storing metrics from
+                          platform components running on fleet managed clusters.
+                        properties:
+                          collection:
+                            description: |-
+                              Collection defines the spec for the addon to collect and forward metrics
+                              from fleet managed clusters.
+                            properties:
+                              enabled:
+                                description: Enabled defines a flag to enable/disable
+                                  the platform metrics collection.
+                                type: boolean
+                            type: object
+                        type: object
                     type: object
                   userWorkloads:
                     description: |-
@@ -10050,6 +10066,22 @@ spec:
                                 type: object
                             type: object
                         type: object
+                      metrics:
+                        description: |-
+                          Metrics defines the spec for the addon to collect, forward and store metrics
+                          from user workloads hosted on fleet managed clusters.
+                        properties:
+                          collection:
+                            description: |-
+                              Collection defines the spec for the addon to collect and forward metrics
+                              from user workloads hosted on fleet managed clusters.
+                            properties:
+                              enabled:
+                                description: Enabled defines a flag to enable/disable
+                                  the platform metrics collection.
+                                type: boolean
+                            type: object
+                        type: object
                       traces:
                         description: |-
                           Traces defines the spec for the addon to collect, forward and store traces

diff --git a/...servability/controllers/multiclusterobservability/multiclusterobservability_controller.go b/...servability/controllers/multiclusterobservability/multiclusterobservability_controller.go
@@ -52,6 +52,7 @@ import (
 	placementctrl "github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/controllers/placementrule"
 	certctrl "github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/pkg/certificates"
 	"github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/pkg/config"
+	mcoconfig "github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/pkg/config"
 	"github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/pkg/rendering"
 	smctrl "github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/pkg/servicemonitor"
 	"github.com/stolostron/multicluster-observability-operator/operators/multiclusterobservability/pkg/util"
@@ -266,10 +267,16 @@ func (r *MultiClusterObservabilityReconciler) Reconcile(ctx context.Context, req
 	}
 	disableMCOACMAORender := !apierrors.IsNotFound(err)
 
+	obsAPIURL, err := mcoconfig.GetObsAPIExternalURL(ctx, r.Client, mcoconfig.GetDefaultNamespace())
+	if err != nil {
+		return ctrl.Result{}, fmt.Errorf("failed to get the Observatorium API URL: %w", err)
+	}
+
 	// Build render options
 	rendererOptions := &rendering.RendererOptions{
 		MCOAOptions: rendering.MCOARendererOptions{
-			DisableCMAORender: disableMCOACMAORender,
+			DisableCMAORender:  disableMCOACMAORender,
+			MetricsHubHostname: obsAPIURL.Host,
 		},
 	}
 

diff --git a/...servability/manifests/base/multicluster-observability-addon/cluster_management_addon.yaml b/...servability/manifests/base/multicluster-observability-addon/cluster_management_addon.yaml
@@ -29,21 +29,16 @@ spec:
     # Describes the default Instrumentation type applied to all managed clusters.
     - group: opentelemetry.io
       resource: instrumentations
+    # Metrics forwarding configuration types.
+    - group: monitoring.coreos.com
+      resource: prometheusagents
+    - group: monitoring.coreos.com
+      resource: scrapeconfigs
+    - group: monitoring.coreos.com
+      resource: prometheusrules
   installStrategy:
     type: Placements
     placements:
       - name: global
         namespace: open-cluster-management-global-set
-        configs:
-          - group: observability.openshift.io
-            resource: clusterlogforwarders
-            name: instance
-            namespace: open-cluster-management-observability
-          - group: opentelemetry.io
-            resource: opentelemetrycollectors
-            name: instance
-            namespace: open-cluster-management-observability
-          - group: opentelemetry.io
-            resource: instrumentations
-            name: instance
-            namespace: open-cluster-management-observability
+        configs: []
diff --git a/...lticlusterobservability/manifests/base/multicluster-observability-addon/cluster_role.yaml b/...lticlusterobservability/manifests/base/multicluster-observability-addon/cluster_role.yaml
@@ -68,3 +68,7 @@
     - apiGroups: ["opentelemetry.io"]
       resources: ["opentelemetrycollectors", "instrumentations"]
       verbs: ["get", "list", "watch"]
+    # Role for addon to perform prometheus specific actions
+    - apiGroups: ["monitoring.coreos.com"]
+      resources: ["prometheusagents", "scrapeconfigs", "prometheusrules"]
+      verbs: ["get", "list", "watch", "create", "update", "delete"]
diff --git a/operators/multiclusterobservability/pkg/config/config.go b/operators/multiclusterobservability/pkg/config/config.go
@@ -230,13 +230,15 @@ const (
 	ClusterLogForwarderCRDName    = "clusterlogforwarders.observability.openshift.io"
 	OpenTelemetryCollectorCRDName = "opentelemetrycollectors.opentelemetry.io"
 	InstrumentationCRDName        = "instrumentations.opentelemetry.io"
+	PrometheusAgentCRDName        = "prometheusagents.monitoring.coreos.com"
 )
 
 var (
 	mcoaSupportedCRDs = map[string]string{
 		ClusterLogForwarderCRDName:    "v1",
 		OpenTelemetryCollectorCRDName: "v1beta1",
 		InstrumentationCRDName:        "v1alpha1",
+		PrometheusAgentCRDName:        "v1alpha1",
 	}
 )
 

diff --git a/operators/multiclusterobservability/pkg/config/config_test.go b/operators/multiclusterobservability/pkg/config/config_test.go
@@ -809,6 +809,7 @@ func TestGetMCOASupportedCRDNames(t *testing.T) {
 		"clusterlogforwarders.observability.openshift.io",
 		"opentelemetrycollectors.opentelemetry.io",
 		"instrumentations.opentelemetry.io",
+		"prometheusagents.monitoring.coreos.com",
 	}
 
 	result := GetMCOASupportedCRDNames()