Tag Prometheus run time info metrics as deprecated (#1217)

Deprecate run time info metrics that provide node name and pod name
from Antrea 0.10.0. These metrics will be removed Antrea 0.11.0.

In addition, antrea-prometheus.yaml is changed to add nodename and
pod name to instance label instead of default IP:port. This will help
with promql queries to search for metrics on a specific node or pod.

build/yamls/antrea-prometheus.yml

@@ -89,6 +89,8 @@ data:
         - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_pod_container_name]
           action: keep
           regex: kube-system;antrea-controller
+        - source_labels: [__meta_kubernetes_pod_node_name, __meta_kubernetes_pod_name]
+          target_label: instance
 
     # Scrape Antrea Agents metrics
       - job_name: 'antrea-agents'
@@ -103,6 +105,8 @@ data:
         - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_pod_container_name]
           action: keep
           regex: kube-system;antrea-agent
+        - source_labels: [__meta_kubernetes_pod_node_name, __meta_kubernetes_pod_name]
+          target_label: instance
 ---
 # Prometheus Server deployment
 apiVersion: apps/v1

docs/prometheus-integration.md

@@ -69,7 +69,10 @@ rules:
 
 ### Antrea Components Scraping configuration
 Add the following jobs to Prometheus scraping configuration to enable metrics
-collection from Antrea components
+collection from Antrea components. Antrea Agent metrics endpoint is exposed through
+Antrea apiserver on `apiport` config parameter given in `antrea-agent.conf` (default
+value is 10350). Antrea Controller metrics endpoint is exposed through Antrea apiserver
+on `apiport` config parameter given in `antrea-controller.conf` (default value is 10349). 
 
 #### Controller Scraping
 ```yaml
@@ -85,6 +88,8 @@ relabel_configs:
 - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_pod_container_name]
   action: keep
   regex: kube-system;antrea-controller
+- source_labels: [__meta_kubernetes_pod_node_name, __meta_kubernetes_pod_name]
+  target_label: instance
 ```
 
 #### Agent Scraping
@@ -101,6 +106,8 @@ relabel_configs:
 - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_pod_container_name]
   action: keep
   regex: kube-system;antrea-agent
+- source_labels: [__meta_kubernetes_pod_node_name, __meta_kubernetes_pod_name]
+  target_label: instance
 ```
 For further reference see the enclosed 
 [configuration file](/build/yamls/antrea-prometheus.yml).

pkg/agent/metrics/prometheus.go

@@ -103,20 +103,19 @@ func InitializePrometheusMetrics() {
 	if err != nil {
 		klog.Errorf("Failed to retrieve agent K8S node name: %v", err)
 	}
-
-	gaugeHost := metrics.NewGauge(&metrics.GaugeOpts{
-		Name: "antrea_agent_runtime_info",
-		Help: "Antrea agent runtime info , defined as labels. The value of the gauge is always set to 1.",
-
-		ConstLabels:    metrics.Labels{"k8s_nodename": nodeName, "k8s_podname": env.GetPodName()},
-		StabilityLevel: metrics.STABLE,
+	deprecatedGaugeHost := metrics.NewGauge(&metrics.GaugeOpts{
+		Name:              "antrea_agent_runtime_info",
+		Help:              "Antrea agent runtime info (Deprecated since Antrea 0.10.0), defined as labels. The value of the gauge is always set to 1.",
+		ConstLabels:       metrics.Labels{"k8s_nodename": nodeName, "k8s_podname": env.GetPodName()},
+		StabilityLevel:    metrics.STABLE,
+		DeprecatedVersion: "0.10.0",
 	})
-	if err := legacyregistry.Register(gaugeHost); err != nil {
+	if err := legacyregistry.Register(deprecatedGaugeHost); err != nil {
 		klog.Error("Failed to register antrea_agent_runtime_info with Prometheus")
 	}
 	// This must be after registering the metrics.Gauge as it is lazily instantiated
 	// and will not measure anything unless the collector is first registered.
-	gaugeHost.Set(1)
+	deprecatedGaugeHost.Set(1)
 
 	InitializePodMetrics()
 	InitializeNetworkPolicyMetrics()

pkg/controller/metrics/prometheus.go

@@ -72,24 +72,26 @@ var (
 
 // Initialize Prometheus metrics collection.
 func InitializePrometheusMetrics() {
+	klog.Info("Initializing prometheus metrics")
+
 	nodeName, err := env.GetNodeName()
 	if err != nil {
 		klog.Errorf("Failed to retrieve controller K8S node name: %v", err)
 	}
-
-	klog.Info("Initializing prometheus metrics")
-	gaugeHost := metrics.NewGauge(&metrics.GaugeOpts{
-		Name:           "antrea_controller_runtime_info",
-		Help:           "Antrea controller runtime info, defined as labels. The value of the gauge is always set to 1.",
-		ConstLabels:    metrics.Labels{"k8s_nodename": nodeName, "k8s_podname": env.GetPodName()},
-		StabilityLevel: metrics.STABLE,
+	deprecatedGaugeHost := metrics.NewGauge(&metrics.GaugeOpts{
+		Name:              "antrea_controller_runtime_info",
+		Help:              "Antrea controller runtime info (Deprecated since Antrea 0.10.0), defined as labels. The value of the gauge is always set to 1.",
+		ConstLabels:       metrics.Labels{"k8s_nodename": nodeName, "k8s_podname": env.GetPodName()},
+		StabilityLevel:    metrics.STABLE,
+		DeprecatedVersion: "0.10.0",
 	})
-	if err = legacyregistry.Register(gaugeHost); err != nil {
+	if err = legacyregistry.Register(deprecatedGaugeHost); err != nil {
 		klog.Errorf("Failed to register antrea_controller_runtime_info with Prometheus: %s", err.Error())
 	}
 	// This must be after registering the metrics.Gauge as it is lazily instantiated
 	// and will not measure anything unless the collector is first registered.
-	gaugeHost.Set(1)
+	deprecatedGaugeHost.Set(1)
+
 	if err := legacyregistry.Register(OpsAppliedToGroupProcessed); err != nil {
 		klog.Errorf("Failed to register antrea_controller_applied_to_group_processed with Prometheus: %s", err.Error())
 	}