Implement Prometheus instrumentation (#99)

* Add Prometheus exporter for mesh metrics
- appmesh_mesh_state gauge
- appmesh_virtual_node_state gauge
- appmesh_virtual_service_state gauge
- appmesh_api_request_duration_seconds histogram
* Add Prometheus instrumentation to App Mesh API client
Records the duration of App Mesh API calls based on object kind, name and operation type. The operation type can be get, create, update or delete. The object kind can be mesh, virtual node, virtual route, virtual router or virtual service.
* Add Prometheus instrumentation to controller
Record mesh, virtual node and virtual service operations as gauges. For each object the gauge value represents the current state, 1 means that the object is active while 0 means that the object has been deleted.
* Add tests for mesh metrics recorder

Signed-off-by: stefanprodan <stefan.prodan@gmail.com>

cmd/app-mesh-controller/root.go

@@ -15,16 +15,11 @@ import (
 	"k8s.io/client-go/tools/clientcmd"
 	"k8s.io/klog"
 
-	// TODO(nic) Don't depend on k8s.io/kubernetes, just duplicate the logic in this package -- it will be a
-	// smaller headache.
-	//_ "k8s.io/kubernetes/pkg/client/metrics/prometheus" // for client metric registration
-	//_ "k8s.io/kubernetes/pkg/util/reflector/prometheus" // for reflector metric registration
-	//_ "k8s.io/kubernetes/pkg/util/workqueue/prometheus" // for workqueue metric registration
-
 	"github.com/aws/aws-app-mesh-controller-for-k8s/pkg/aws"
 	meshclientset "github.com/aws/aws-app-mesh-controller-for-k8s/pkg/client/clientset/versioned"
 	meshinformers "github.com/aws/aws-app-mesh-controller-for-k8s/pkg/client/informers/externalversions"
 	"github.com/aws/aws-app-mesh-controller-for-k8s/pkg/controller"
+	"github.com/aws/aws-app-mesh-controller-for-k8s/pkg/metrics"
 )
 
 var (
@@ -89,7 +84,8 @@ var rootCmd = &cobra.Command{
 			klog.Fatal(err)
 		}
 
-		cloud, err := aws.NewCloud(cfg.aws)
+		stats := metrics.NewRecorder(true)
+		cloud, err := aws.NewCloud(cfg.aws, stats)
 		if err != nil {
 			klog.Fatal(err)
 		}
@@ -117,6 +113,7 @@ var rootCmd = &cobra.Command{
 			meshInformerFactory.Appmesh().V1beta1().Meshes(),
 			meshInformerFactory.Appmesh().V1beta1().VirtualNodes(),
 			meshInformerFactory.Appmesh().V1beta1().VirtualServices(),
+			stats,
 		)
 
 		if err != nil {

go.mod

@@ -10,6 +10,7 @@ require (
 	github.com/imdario/mergo v0.3.7 // indirect
 	github.com/inconshreveable/mousetrap v1.0.0 // indirect
 	github.com/prometheus/client_golang v0.9.2
+	github.com/prometheus/client_model v0.0.0-20180712105110-5c3871d89910
 	github.com/spf13/cobra v0.0.3
 	github.com/spf13/pflag v1.0.5
 	github.com/spf13/viper v1.3.1

pkg/aws/appmesh.go

@@ -76,6 +76,11 @@ func (v *Mesh) Name() string {
 
 // GetMesh calls describe mesh.
 func (c *Cloud) GetMesh(ctx context.Context, name string) (*Mesh, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("mesh", name, "get", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DescribeMeshTimeout)
 	defer cancel()
 
@@ -96,6 +101,11 @@ func (c *Cloud) GetMesh(ctx context.Context, name string) (*Mesh, error) {
 
 // CreateMesh converts the desired mesh spec into CreateMeshInput and calls create mesh.
 func (c *Cloud) CreateMesh(ctx context.Context, mesh *appmeshv1beta1.Mesh) (*Mesh, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("mesh", mesh.Name, "create", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*CreateMeshTimeout)
 	defer cancel()
 
@@ -116,6 +126,11 @@ func (c *Cloud) CreateMesh(ctx context.Context, mesh *appmeshv1beta1.Mesh) (*Mes
 
 // DeleteMesh deletes the given mesh
 func (c *Cloud) DeleteMesh(ctx context.Context, name string) (*Mesh, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("mesh", name, "delete", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DeleteMeshTimeout)
 	defer cancel()
 
@@ -231,6 +246,11 @@ func (v *VirtualNode) BackendsSet() set.Set {
 
 // GetVirtualNode calls describe virtual node.
 func (c *Cloud) GetVirtualNode(ctx context.Context, name string, meshName string) (*VirtualNode, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_node", name, "get", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DescribeVirtualNodeTimeout)
 	defer cancel()
 
@@ -253,6 +273,11 @@ func (c *Cloud) GetVirtualNode(ctx context.Context, name string, meshName string
 // CreateVirtualNode converts the desired virtual node spec into CreateVirtualNodeInput and calls create
 // virtual node.
 func (c *Cloud) CreateVirtualNode(ctx context.Context, vnode *appmeshv1beta1.VirtualNode) (*VirtualNode, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_node", vnode.Name, "create", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*CreateVirtualNodeTimeout)
 	defer cancel()
 
@@ -341,6 +366,11 @@ func (c *Cloud) CreateVirtualNode(ctx context.Context, vnode *appmeshv1beta1.Vir
 // UpdateVirtualNode converts the desired virtual node spec into UpdateVirtualNodeInput and calls update
 // virtual node.
 func (c *Cloud) UpdateVirtualNode(ctx context.Context, vnode *appmeshv1beta1.VirtualNode) (*VirtualNode, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_node", vnode.Name, "update", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*UpdateVirtualNodeTimeout)
 	defer cancel()
 
@@ -427,6 +457,11 @@ func (c *Cloud) UpdateVirtualNode(ctx context.Context, vnode *appmeshv1beta1.Vir
 }
 
 func (c *Cloud) DeleteVirtualNode(ctx context.Context, name string, meshName string) (*VirtualNode, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_node", name, "delete", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DeleteVirtualNodeTimeout)
 	defer cancel()
 
@@ -476,6 +511,11 @@ func (v *VirtualService) Status() string {
 
 // GetVirtualService calls describe virtual service.
 func (c *Cloud) GetVirtualService(ctx context.Context, name string, meshName string) (*VirtualService, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_service", name, "get", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DescribeVirtualServiceTimeout)
 	defer cancel()
 
@@ -498,6 +538,11 @@ func (c *Cloud) GetVirtualService(ctx context.Context, name string, meshName str
 // CreateVirtualService converts the desired virtual service spec into CreateVirtualServiceInput and calls create
 // virtual service.
 func (c *Cloud) CreateVirtualService(ctx context.Context, vservice *appmeshv1beta1.VirtualService) (*VirtualService, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_service", vservice.Name, "create", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*CreateVirtualServiceTimeout)
 	defer cancel()
 
@@ -526,6 +571,11 @@ func (c *Cloud) CreateVirtualService(ctx context.Context, vservice *appmeshv1bet
 }
 
 func (c *Cloud) UpdateVirtualService(ctx context.Context, vservice *appmeshv1beta1.VirtualService) (*VirtualService, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_service", vservice.Name, "update", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*UpdateVirtualServiceTimeout)
 	defer cancel()
 
@@ -554,6 +604,11 @@ func (c *Cloud) UpdateVirtualService(ctx context.Context, vservice *appmeshv1bet
 }
 
 func (c *Cloud) DeleteVirtualService(ctx context.Context, name string, meshName string) (*VirtualService, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_service", name, "delete", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DeleteVirtualServiceTimeout)
 	defer cancel()
 
@@ -593,6 +648,11 @@ func (v *VirtualRouter) Status() string {
 
 // GetVirtualRouter calls describe virtual router.
 func (c *Cloud) GetVirtualRouter(ctx context.Context, name string, meshName string) (*VirtualRouter, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_router", name, "get", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DescribeVirtualRouterTimeout)
 	defer cancel()
 
@@ -615,6 +675,11 @@ func (c *Cloud) GetVirtualRouter(ctx context.Context, name string, meshName stri
 // CreateVirtualRouter converts the desired virtual service spec into CreateVirtualServiceInput and calls create
 // virtual router.
 func (c *Cloud) CreateVirtualRouter(ctx context.Context, vrouter *appmeshv1beta1.VirtualRouter, meshName string) (*VirtualRouter, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_router", vrouter.Name, "create", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*CreateVirtualRouterTimeout)
 	defer cancel()
 
@@ -652,6 +717,11 @@ func (c *Cloud) CreateVirtualRouter(ctx context.Context, vrouter *appmeshv1beta1
 
 // UpdateVirtualRouter converts the desired virtual router spec into UpdateVirtualRouter calls
 func (c *Cloud) UpdateVirtualRouter(ctx context.Context, vrouter *appmeshv1beta1.VirtualRouter, meshName string) (*VirtualRouter, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_router", vrouter.Name, "update", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*UpdateVirtualRouterTimeout)
 	defer cancel()
 
@@ -688,6 +758,11 @@ func (c *Cloud) UpdateVirtualRouter(ctx context.Context, vrouter *appmeshv1beta1
 }
 
 func (c *Cloud) DeleteVirtualRouter(ctx context.Context, name string, meshName string) (*VirtualRouter, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_router", name, "delete", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DeleteVirtualRouterTimeout)
 	defer cancel()
 
@@ -948,6 +1023,11 @@ func (r Routes) RouteByName(name string) Route {
 
 // GetRoute calls describe route.
 func (c *Cloud) GetRoute(ctx context.Context, name string, routerName string, meshName string) (*Route, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_route", name, "get", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DescribeRouteTimeout)
 	defer cancel()
 
@@ -970,6 +1050,11 @@ func (c *Cloud) GetRoute(ctx context.Context, name string, routerName string, me
 
 // CreateRoute converts the desired virtual service spec into CreateVirtualServiceInput and calls create route.
 func (c *Cloud) CreateRoute(ctx context.Context, route *appmeshv1beta1.Route, routerName string, meshName string) (*Route, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_route", route.Name, "create", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*CreateRouteTimeout)
 	defer cancel()
 
@@ -992,6 +1077,11 @@ func (c *Cloud) CreateRoute(ctx context.Context, route *appmeshv1beta1.Route, ro
 }
 
 func (c *Cloud) GetRoutesForVirtualRouter(ctx context.Context, routerName string, meshName string) (Routes, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_router", routerName, "get", time.Since(begin))
+	}()
+
 	listctx, cancel := context.WithTimeout(ctx, time.Second*ListRoutesTimeout)
 	defer cancel()
 
@@ -1025,6 +1115,11 @@ func (c *Cloud) GetRoutesForVirtualRouter(ctx context.Context, routerName string
 
 // UpdateRoute converts the desired virtual service spec into UpdateRouteInput and calls update route.
 func (c *Cloud) UpdateRoute(ctx context.Context, route *appmeshv1beta1.Route, routerName string, meshName string) (*Route, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_route", route.Name, "update", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*UpdateRouteTimeout)
 	defer cancel()
 
@@ -1047,6 +1142,11 @@ func (c *Cloud) UpdateRoute(ctx context.Context, route *appmeshv1beta1.Route, ro
 }
 
 func (c *Cloud) DeleteRoute(ctx context.Context, name string, routerName string, meshName string) (*Route, error) {
+	begin := time.Now()
+	defer func() {
+		c.stats.SetRequestDuration("virtual_route", name, "delete", time.Since(begin))
+	}()
+
 	ctx, cancel := context.WithTimeout(ctx, time.Second*DeleteRouteTimeout)
 	defer cancel()
 

pkg/aws/cloud.go

@@ -4,9 +4,9 @@ import (
 	"fmt"
 	"time"
 
-	"github.com/aws/aws-sdk-go/aws/ec2metadata"
-
+	"github.com/aws/aws-app-mesh-controller-for-k8s/pkg/metrics"
 	"github.com/aws/aws-sdk-go/aws"
+	"github.com/aws/aws-sdk-go/aws/ec2metadata"
 	"github.com/aws/aws-sdk-go/aws/session"
 	"github.com/aws/aws-sdk-go/service/appmesh"
 	"github.com/aws/aws-sdk-go/service/appmesh/appmeshiface"
@@ -28,6 +28,8 @@ type Cloud struct {
 
 	namespaceIDCache cache.Store
 	serviceIDCache   cache.Store
+
+	stats *metrics.Recorder
 }
 
 type cloudmapServiceCacheItem struct {
@@ -50,7 +52,7 @@ type CloudMapNamespaceSummary struct {
 	NamespaceType string
 }
 
-func NewCloud(opts CloudOptions) (CloudAPI, error) {
+func NewCloud(opts CloudOptions, stats *metrics.Recorder) (CloudAPI, error) {
 	cfg := &aws.Config{Region: aws.String(opts.Region)}
 
 	session, err := session.NewSession(cfg)
@@ -77,5 +79,6 @@ func NewCloud(opts CloudOptions) (CloudAPI, error) {
 		serviceIDCache: cache.NewTTLStore(func(obj interface{}) (string, error) {
 			return obj.(*cloudmapServiceCacheItem).key, nil
 		}, 60*time.Second),
+		stats: stats,
 	}, nil
 }

pkg/controller/controller.go

@@ -11,6 +11,7 @@ import (
 	meshscheme "github.com/aws/aws-app-mesh-controller-for-k8s/pkg/client/clientset/versioned/scheme"
 	meshinformers "github.com/aws/aws-app-mesh-controller-for-k8s/pkg/client/informers/externalversions/appmesh/v1beta1"
 	meshlisters "github.com/aws/aws-app-mesh-controller-for-k8s/pkg/client/listers/appmesh/v1beta1"
+	"github.com/aws/aws-app-mesh-controller-for-k8s/pkg/metrics"
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/util/runtime"
 	utilruntime "k8s.io/apimachinery/pkg/util/runtime"
@@ -69,6 +70,9 @@ type Controller struct {
 	// recorder is an event recorder for recording Event resources to the
 	// Kubernetes API.
 	recorder record.EventRecorder
+
+	// stats records mesh Prometheus metrics
+	stats *metrics.Recorder
 }
 
 func NewController(
@@ -78,7 +82,8 @@ func NewController(
 	podInformer coreinformers.PodInformer,
 	meshInformer meshinformers.MeshInformer,
 	virtualNodeInformer meshinformers.VirtualNodeInformer,
-	virtualServiceInformer meshinformers.VirtualServiceInformer) (*Controller, error) {
+	virtualServiceInformer meshinformers.VirtualServiceInformer,
+	stats *metrics.Recorder) (*Controller, error) {
 
 	utilruntime.Must(meshscheme.AddToScheme(scheme.Scheme))
 	klog.V(4).Info("Creating event broadcaster")
@@ -105,6 +110,7 @@ func NewController(
 		sq:                   workqueue.NewRateLimitingQueue(workqueue.DefaultControllerRateLimiter()),
 		pq:                   workqueue.NewRateLimitingQueue(workqueue.DefaultControllerRateLimiter()),
 		recorder:             recorder,
+		stats:                stats,
 	}
 
 	podInformer.Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{

pkg/controller/mesh.go

@@ -35,6 +35,7 @@ func (c *Controller) handleMesh(key string) error {
 	// Resources with finalizers are not deleted immediately,
 	// instead the deletion timestamp is set when a client deletes them.
 	if !mesh.DeletionTimestamp.IsZero() {
+		c.stats.SetMeshInactive(mesh.Name)
 		// Resource is being deleted, process finalizers
 		return c.handleMeshDelete(ctx, mesh)
 	}
@@ -65,6 +66,8 @@ func (c *Controller) handleMesh(key string) error {
 		}
 	}
 
+	c.stats.SetMeshActive(mesh.Name)
+
 	return nil
 }
 

pkg/controller/virtualnode.go

@@ -52,6 +52,7 @@ func (c *Controller) handleVNode(key string) error {
 	// Resources with finalizers are not deleted immediately,
 	// instead the deletion timestamp is set when a client deletes them.
 	if !vnode.DeletionTimestamp.IsZero() {
+		c.stats.SetVirtualNodeInactive(vnode.Name, vnode.Spec.MeshName)
 		// Resource is being deleted, process finalizers
 		return c.handleVNodeDelete(ctx, vnode, copy)
 	}
@@ -108,6 +109,8 @@ func (c *Controller) handleVNode(key string) error {
 		}
 	}
 
+	c.stats.SetVirtualNodeActive(vnode.Name, vnode.Spec.MeshName)
+
 	updated, err := c.updateVNodeStatus(copy, targetNode)
 	if err != nil {
 		return fmt.Errorf("error updating virtual node status: %s", err)