Antrea Prometheus integration (#236)

Integrate with Prometheus monitoring solution. Integration of the Prometheus client into Antrea controller and agent allows the exposure of various metrics to Prometheus server. In addition to Antrea's own set of metrics, Prometheus client will also expose metrics which are defined by various components which are part of the Antrea ecosystem, e.g golang, Prometheus itself etc.
antrea-io · Mar 22, 2020 · e15a3e6 · e15a3e6
1 parent 938169e
commit e15a3e6
Show file tree

Hide file tree

Showing 12 changed files with 279 additions and 13 deletions.
diff --git a/build/yamls/antrea-ipsec.yml b/build/yamls/antrea-ipsec.yml
@@ -239,7 +239,7 @@ subjects:
 ---
 apiVersion: v1
 data:
-  antrea-agent.conf: |
+  antrea-agent.conf: |-
     # Name of the OpenVSwitch bridge antrea-agent will create and use.
     # Make sure it doesn't conflict with your existing OpenVSwitch bridges.
     #ovsBridge: br-int
@@ -281,6 +281,15 @@ data:
     #          Underlying network must be capable of supporting Pod traffic across IP subnet.
     # hybrid: noEncap if worker Nodes on same subnet, otherwise encap.
     #trafficEncapMode: encap
+
+    # Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener
+    #enablePrometheusMetrics: false
+
+    # Enable golang metrics exposure via Prometheus.
+    #enablePrometheusGoMetrics: false
+
+    # Enable process metrics exposure via Prometheus.
+    #enablePrometheusProcessMetrics: false
   antrea-cni.conf: |
     {
         "cniVersion":"0.3.0",
@@ -290,13 +299,21 @@ data:
             "type": "host-local"
         }
     }
-  antrea-controller.conf: ""
+  antrea-controller.conf: |-
+    # Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener.
+    #enablePrometheusMetrics: false
+
+    # Enable golang metrics exposure via Prometheus.
+    #enablePrometheusGoMetrics: false
+
+    # Enable process metrics exposure via Prometheus.
+    #enablePrometheusProcessMetrics: false
 kind: ConfigMap
 metadata:
   annotations: {}
   labels:
     app: antrea
-  name: antrea-config-ghc6hct4mg
+  name: antrea-config-ffmk8fd5g4
   namespace: kube-system
 ---
 apiVersion: v1
@@ -311,6 +328,10 @@ type: Opaque
 apiVersion: v1
 kind: Service
 metadata:
+  annotations:
+    prometheus.io/port: "443"
+    prometheus.io/scheme: https
+    prometheus.io/scrape: "true"
   labels:
     app: antrea
   name: antrea
@@ -398,7 +419,7 @@ spec:
         key: node-role.kubernetes.io/master
       volumes:
       - configMap:
-          name: antrea-config-ghc6hct4mg
+          name: antrea-config-ffmk8fd5g4
         name: antrea-config
 ---
 apiVersion: apiregistration.k8s.io/v1
@@ -432,6 +453,10 @@ spec:
       component: antrea-agent
   template:
     metadata:
+      annotations:
+        prometheus.io/port: "10443"
+        prometheus.io/scheme: https
+        prometheus.io/scrape: "true"
       labels:
         app: antrea
         component: antrea-agent
@@ -598,7 +623,7 @@ spec:
         operator: Exists
       volumes:
       - configMap:
-          name: antrea-config-ghc6hct4mg
+          name: antrea-config-ffmk8fd5g4
         name: antrea-config
       - hostPath:
           path: /etc/cni/net.d

diff --git a/build/yamls/antrea.yml b/build/yamls/antrea.yml
@@ -239,7 +239,7 @@ subjects:
 ---
 apiVersion: v1
 data:
-  antrea-agent.conf: |
+  antrea-agent.conf: |-
     # Name of the OpenVSwitch bridge antrea-agent will create and use.
     # Make sure it doesn't conflict with your existing OpenVSwitch bridges.
     #ovsBridge: br-int
@@ -281,6 +281,15 @@ data:
     #          Underlying network must be capable of supporting Pod traffic across IP subnet.
     # hybrid: noEncap if worker Nodes on same subnet, otherwise encap.
     #trafficEncapMode: encap
+
+    # Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener
+    #enablePrometheusMetrics: false
+
+    # Enable golang metrics exposure via Prometheus.
+    #enablePrometheusGoMetrics: false
+
+    # Enable process metrics exposure via Prometheus.
+    #enablePrometheusProcessMetrics: false
   antrea-cni.conf: |
     {
         "cniVersion":"0.3.0",
@@ -290,18 +299,30 @@ data:
             "type": "host-local"
         }
     }
-  antrea-controller.conf: ""
+  antrea-controller.conf: |-
+    # Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener.
+    #enablePrometheusMetrics: false
+
+    # Enable golang metrics exposure via Prometheus.
+    #enablePrometheusGoMetrics: false
+
+    # Enable process metrics exposure via Prometheus.
+    #enablePrometheusProcessMetrics: false
 kind: ConfigMap
 metadata:
   annotations: {}
   labels:
     app: antrea
-  name: antrea-config-4gb24b784b
+  name: antrea-config-k49g4578m4
   namespace: kube-system
 ---
 apiVersion: v1
 kind: Service
 metadata:
+  annotations:
+    prometheus.io/port: "443"
+    prometheus.io/scheme: https
+    prometheus.io/scrape: "true"
   labels:
     app: antrea
   name: antrea
@@ -389,7 +410,7 @@ spec:
         key: node-role.kubernetes.io/master
       volumes:
       - configMap:
-          name: antrea-config-4gb24b784b
+          name: antrea-config-k49g4578m4
         name: antrea-config
 ---
 apiVersion: apiregistration.k8s.io/v1
@@ -423,6 +444,10 @@ spec:
       component: antrea-agent
   template:
     metadata:
+      annotations:
+        prometheus.io/port: "10443"
+        prometheus.io/scheme: https
+        prometheus.io/scrape: "true"
       labels:
         app: antrea
         component: antrea-agent
@@ -557,7 +582,7 @@ spec:
         operator: Exists
       volumes:
       - configMap:
-          name: antrea-config-4gb24b784b
+          name: antrea-config-k49g4578m4
         name: antrea-config
       - hostPath:
           path: /etc/cni/net.d

diff --git a/build/yamls/base/agent.yml b/build/yamls/base/agent.yml
@@ -13,6 +13,10 @@ spec:
     type: RollingUpdate
   template:
     metadata:
+      annotations:
+        prometheus.io/port: "10443"
+        prometheus.io/scrape: "true"
+        prometheus.io/scheme: "https"
       labels:
         component: antrea-agent
     spec:

diff --git a/build/yamls/base/conf/antrea-agent.conf b/build/yamls/base/conf/antrea-agent.conf
@@ -39,3 +39,12 @@
 #          Underlying network must be capable of supporting Pod traffic across IP subnet.
 # hybrid: noEncap if worker Nodes on same subnet, otherwise encap.
 #trafficEncapMode: encap
+
+# Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener
+#enablePrometheusMetrics: false
+
+# Enable golang metrics exposure via Prometheus.
+#enablePrometheusGoMetrics: false
+
+# Enable process metrics exposure via Prometheus.
+#enablePrometheusProcessMetrics: false
diff --git a/build/yamls/base/conf/antrea-controller.conf b/build/yamls/base/conf/antrea-controller.conf
@@ -0,0 +1,8 @@
+# Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener.
+#enablePrometheusMetrics: false
+
+# Enable golang metrics exposure via Prometheus.
+#enablePrometheusGoMetrics: false
+
+# Enable process metrics exposure via Prometheus.
+#enablePrometheusProcessMetrics: false
diff --git a/build/yamls/base/controller.yml b/build/yamls/base/controller.yml
@@ -2,6 +2,10 @@
 apiVersion: v1
 kind: Service
 metadata:
+  annotations:
+    prometheus.io/port: "443"
+    prometheus.io/scrape: "true"
+    prometheus.io/scheme: "https"
   name: antrea
 spec:
   ports:

diff --git a/cmd/antrea-agent/agent.go b/cmd/antrea-agent/agent.go
@@ -30,6 +30,7 @@ import (
 	"github.com/vmware-tanzu/antrea/pkg/agent/controller/networkpolicy"
 	"github.com/vmware-tanzu/antrea/pkg/agent/controller/noderoute"
 	"github.com/vmware-tanzu/antrea/pkg/agent/interfacestore"
+	"github.com/vmware-tanzu/antrea/pkg/agent/metrics"
 	"github.com/vmware-tanzu/antrea/pkg/agent/openflow"
 	"github.com/vmware-tanzu/antrea/pkg/agent/route"
 	"github.com/vmware-tanzu/antrea/pkg/apis/networking/v1beta1"
@@ -146,6 +147,13 @@ func run(o *Options) error {
 
 	go networkPolicyController.Run(stopCh)
 
+	if o.config.EnablePrometheusMetrics {
+		go metrics.StartListener(
+			o.config.EnablePrometheusGoMetrics,
+			o.config.EnablePrometheusProcessMetrics,
+			o.config.OVSBridge, ifaceStore, ofClient)
+	}
+
 	agentMonitor := monitor.NewAgentMonitor(
 		crdClient,
 		o.config.OVSBridge,

diff --git a/cmd/antrea-agent/config.go b/cmd/antrea-agent/config.go
@@ -71,4 +71,13 @@ type AgentConfig struct {
 	//          Underlying network must be capable of supporting Pod traffic across IP subnet.
 	// Hybrid: noEncap if worker Nodes on same subnet, otherwise encap.
 	TrafficEncapMode string `yaml:"trafficEncapMode,omitempty"`
+	// Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener
+	// Defaults to false.
+	EnablePrometheusMetrics bool `yaml:"enablePrometheusMetrics,omitempty"`
+	// Enable golang metrics exposure via Prometheus
+	// Defaults to false.
+	EnablePrometheusGoMetrics bool `yaml:"enablePrometheusGoMetrics,omitempty"`
+	// Enable process metrics exposure via Prometheus
+	// Defaults to false.
+	EnablePrometheusProcessMetrics bool `yaml:"enablePrometheusProcessMetrics,omitempty"`
 }
diff --git a/cmd/antrea-controller/config.go b/cmd/antrea-controller/config.go
@@ -22,4 +22,13 @@ type ControllerConfig struct {
 	// clientConnection specifies the kubeconfig file and client connection settings for the agent
 	// to communicate with the apiserver.
 	ClientConnection componentbaseconfig.ClientConnectionConfiguration `yaml:"clientConnection"`
+	// Enable metrics exposure via Prometheus. Initializes Prometheus metrics listener
+	// Defaults to false.
+	EnablePrometheusMetrics bool `yaml:"enablePrometheusMetrics,omitempty"`
+	// Enable golang metrics exposure via Prometheus
+	// Defaults to false.
+	EnablePrometheusGoMetrics bool `yaml:"enablePrometheusGoMetrics,omitempty"`
+	// Enable process metrics exposure via Prometheus
+	// Defaults to false.
+	EnablePrometheusProcessMetrics bool `yaml:"enablePrometheusProcessMetrics,omitempty"`
 }
diff --git a/cmd/antrea-controller/controller.go b/cmd/antrea-controller/controller.go
@@ -17,8 +17,12 @@ package main
 import (
 	"fmt"
 	"net"
+	"net/http"
+	"os"
 	"time"
 
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
 	genericopenapi "k8s.io/apiserver/pkg/endpoints/openapi"
 	genericapiserver "k8s.io/apiserver/pkg/server"
 	genericoptions "k8s.io/apiserver/pkg/server/options"
@@ -71,7 +75,8 @@ func run(o *Options) error {
 	apiServerConfig, err := createAPIServerConfig(o.config.ClientConnection.Kubeconfig,
 		addressGroupStore,
 		appliedToGroupStore,
-		networkPolicyStore)
+		networkPolicyStore,
+		o.config.EnablePrometheusMetrics)
 	if err != nil {
 		return fmt.Errorf("error creating API server config: %v", err)
 	}
@@ -94,22 +99,62 @@ func run(o *Options) error {
 
 	go apiServer.GenericAPIServer.PrepareRun().Run(stopCh)
 
+	if o.config.EnablePrometheusMetrics {
+		go createPrometheusMetricsListener(
+			o.config.EnablePrometheusGoMetrics,
+			o.config.EnablePrometheusGoMetrics)
+	}
+
 	<-stopCh
 	klog.Info("Stopping Antrea controller")
 	return nil
 }
 
+// Initialize Prometheus listener and metrics collection.
+func createPrometheusMetricsListener(
+	enablePrometheusGoMetrics bool,
+	enablePrometheusProcessMetrics bool) {
+	hostname, err := os.Hostname()
+	if err != nil {
+		klog.Errorf("Failed to retrieve agent node name, %v", err)
+	}
+
+	klog.Info("Initializing prometheus")
+	gaugeHost := prometheus.NewGauge(prometheus.GaugeOpts{
+		Name: "antrea_controller_host",
+		Help: "Antrea controller hostname (as a label), typically used in grouping/aggregating stats; " +
+			"the label defaults to the hostname of the host but can be overridden by configuration. " +
+			"The value of the gauge is always set to 1.",
+		ConstLabels: prometheus.Labels{"host": hostname},
+	})
+	gaugeHost.Set(1)
+	prometheus.MustRegister(gaugeHost)
+	http.Handle("/metrics", promhttp.Handler())
+
+	if !enablePrometheusGoMetrics {
+		klog.Info("Golang metrics are disabled")
+		prometheus.Unregister(prometheus.NewGoCollector())
+	}
+	if !enablePrometheusProcessMetrics {
+		klog.Info("Process metrics are disabled")
+		prometheus.Unregister(prometheus.NewProcessCollector(prometheus.ProcessCollectorOpts{}))
+	}
+}
+
 func createAPIServerConfig(kubeconfig string,
 	addressGroupStore storage.Interface,
 	appliedToGroupStore storage.Interface,
-	networkPolicyStore storage.Interface) (*apiserver.Config, error) {
+	networkPolicyStore storage.Interface,
+	enablePrometheusMetrics bool) (*apiserver.Config, error) {
 	// TODO:
 	// 1. Support user-provided certificate.
 	// 2. Support configurable https port.
 	secureServing := genericoptions.NewSecureServingOptions().WithLoopback()
 	authentication := genericoptions.NewDelegatingAuthenticationOptions()
 	authorization := genericoptions.NewDelegatingAuthorizationOptions()
-
+	if enablePrometheusMetrics {
+		authorization.WithAlwaysAllowPaths("/metrics")
+	}
 	// Set the PairName but leave certificate directory blank to generate in-memory by default
 	secureServing.ServerCert.CertDirectory = ""
 	secureServing.ServerCert.PairName = "antrea-apiserver"

diff --git a/go.mod b/go.mod
@@ -25,6 +25,7 @@ require (
 	github.com/imdario/mergo v0.3.7 // indirect
 	github.com/j-keck/arping v1.0.0
 	github.com/kevinburke/ssh_config v0.0.0-20190725054713-01f96b0aa0cd
+	github.com/prometheus/client_golang v0.9.3-0.20190127221311-3c4408c8b829
 	github.com/satori/go.uuid v1.2.0
 	github.com/spf13/cobra v0.0.5
 	github.com/spf13/pflag v1.0.3