Update example to work with old and new Stackdriver resource model

custom-metrics-stackdriver-adapter/.gitignore

@@ -1,4 +1,5 @@
 build
-testing/sd_dummy_exporter
+examples/direct-to-sd/sd_dummy_exporter
+examples/prometheus-to-sd/prometheus_dummy_exporter
 *.swp
 *~

custom-metrics-stackdriver-adapter/examples/direct-to-sd/Makefile

@@ -1,4 +1,4 @@
-TAG = v0.1.0
+TAG = v0.2.0
 PREFIX = gcr.io/google-containers
 
 build: sd_dummy_exporter

custom-metrics-stackdriver-adapter/examples/direct-to-sd/sd_dummy_exporter.go

@@ -33,31 +33,59 @@ import (
 
 // SD Dummy Exporter is a testing utility that exports a metric of constant value to Stackdriver
 // in a loop. Metric name and value can be specified with flags 'metric-name' and 'metric-value'.
-// SD Dummy Exporter assumes that it runs as a pod in GCE or GKE cluster, and the pod id is passed
-// to it with 'pod-id' flag (which can be passed to a pod via Downward API).
+// SD Dummy Exporter assumes that it runs as a pod in GCE or GKE cluster, and the pod id, pod name
+// and namespace are passed to it with 'pod-id', 'pod-name' and 'namespace' flags.
+// Pod ID and pod name can be passed to a pod via Downward API.
 func main() {
 	// Gather pod information
 	podId := flag.String("pod-id", "", "pod id")
+	namespace := flag.String("namespace", "", "namespace")
+	podName := flag.String("pod-name", "", "pod name")
 	metricName := flag.String("metric-name", "foo", "custom metric name")
 	metricValue := flag.Int64("metric-value", 0, "custom metric value")
+	// Whether to use old Stackdriver resource model - use monitored resource "gke_container"
+	// For old resource model, podId flag has to be set.
+	useOldResourceModel := flag.Bool("use-old-resource-model", true, "use old stackdriver resource model")
+	// Whether to use new Stackdriver resource model - use monitored resource "k8s_pod"
+	// For new resource model, podName and namespace flags have to be set.
+	useNewResourceModel := flag.Bool("use-new-resource-model", false, "use new stackdriver resource model")
 	flag.Parse()
 
-	if *podId == "" {
+	if *podId == "" && *useOldResourceModel {
 		log.Fatalf("No pod id specified.")
 	}
 
+	if *podName == "" && *useNewResourceModel {
+		log.Fatalf("No pod name specified.")
+	}
+
+	if *namespace == "" && *useNewResourceModel {
+		log.Fatalf("No pod namespace specified.")
+	}
+
 	stackdriverService, err := getStackDriverService()
 	if err != nil {
 		log.Fatalf("Error getting Stackdriver service: %v", err)
 	}
 
-	labels := getResourceLabels(*podId)
+	oldModelLabels := getResourceLabelsForOldModel(*podId)
+	newModelLabels := getResourceLabelsForNewModel(*namespace, *podName)
 	for {
-		err := exportMetric(stackdriverService, *metricName, *metricValue, labels)
-		if err != nil {
-			log.Printf("Failed to write time series data: %v\n", err)
-		} else {
-			log.Printf("Finished writing time series with value: %v\n", metricValue)
+		if *useOldResourceModel {
+			err := exportMetric(stackdriverService, *metricName, *metricValue, "gke_container", oldModelLabels)
+			if err != nil {
+				log.Printf("Failed to write time series data for old resource model: %v\n", err)
+			} else {
+				log.Printf("Finished writing time series for new resource model with value: %v\n", metricValue)
+			}
+		}
+		if *useNewResourceModel {
+			err := exportMetric(stackdriverService, *metricName, *metricValue, "k8s_pod", newModelLabels)
+			if err != nil {
+				log.Printf("Failed to write time series data for new resource model: %v\n", err)
+			} else {
+				log.Printf("Finished writing time series for new resource model with value: %v\n", metricValue)
+			}
 		}
 		time.Sleep(5000 * time.Millisecond)
 	}
@@ -68,10 +96,10 @@ func getStackDriverService() (*monitoring.Service, error) {
 	return monitoring.New(oauthClient)
 }
 
-// getResourceLabels returns resource labels needed to correctly label metric data
-// exported to StackDriver. Labels contain details on the cluster (name, zone, project id)
-// and pod for which the metric is exported (id)
-func getResourceLabels(podId string) map[string]string {
+// getResourceLabelsForOldModel returns resource labels needed to correctly label metric data
+// exported to StackDriver. Labels contain details on the cluster (project id, name)
+// and pod for which the metric is exported (zone, id).
+func getResourceLabelsForOldModel(podId string) map[string]string {
 	projectId, _ := gce.ProjectID()
 	zone, _ := gce.Zone()
 	clusterName, _ := gce.InstanceAttributeValue("cluster-name")
@@ -90,8 +118,26 @@ func getResourceLabels(podId string) map[string]string {
 	}
 }
 
+// getResourceLabelsForNewModel returns resource labels needed to correctly label metric data
+// exported to StackDriver. Labels contain details on the cluster (project id, location, name)
+// and pod for which the metric is exported (namespace, name).
+func getResourceLabelsForNewModel(namespace, name string) map[string]string {
+	projectId, _ := gce.ProjectID()
+	location, _ := gce.InstanceAttributeValue("cluster-location")
+	location = strings.TrimSpace(location)
+	clusterName, _ := gce.InstanceAttributeValue("cluster-name")
+	clusterName = strings.TrimSpace(clusterName)
+	return map[string]string{
+		"project_id":     projectId,
+		"location":       location,
+		"cluster_name":   clusterName,
+		"namespace_name": namespace,
+		"pod_name":       name,
+	}
+}
+
 func exportMetric(stackdriverService *monitoring.Service, metricName string,
-	metricValue int64, resourceLabels map[string]string) error {
+	metricValue int64, monitoredResource string, resourceLabels map[string]string) error {
 	dataPoint := &monitoring.Point{
 		Interval: &monitoring.TimeInterval{
 			EndTime: time.Now().Format(time.RFC3339),
@@ -108,7 +154,7 @@ func exportMetric(stackdriverService *monitoring.Service, metricName string,
 					Type: "custom.googleapis.com/" + metricName,
 				},
 				Resource: &monitoring.MonitoredResource{
-					Type:   "gke_container",
+					Type:   monitoredResource,
 					Labels: resourceLabels,
 				},
 				Points: []*monitoring.Point{

hack/verify-flags/exceptions.txt

@@ -6,18 +6,27 @@ custom-metrics-stackdriver-adapter/README.md:       podIdFlag := flag.String("po
 custom-metrics-stackdriver-adapter/README.md:     - `pod_id` - set this to pod UID obtained via downward API. Example configuration that passes POD_ID to your application as a flag:
 custom-metrics-stackdriver-adapter/examples/direct-to-sd/sd_dummy_exporter.go:		"namespace_id": "default",
 custom-metrics-stackdriver-adapter/examples/direct-to-sd/sd_dummy_exporter.go:		"pod_id":         podId,
+custom-metrics-stackdriver-adapter/examples/direct-to-sd/sd_dummy_exporter.go:		"pod_name":       name,
 custom-metrics-stackdriver-adapter/examples/direct-to-sd/sd_dummy_exporter.go:		// namespace_id and instance_id don't matter
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:			return timeSeries.Resource.Labels["namespace_name"] + ":" + timeSeries.Resource.Labels["pod_name"], nil
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:		return fmt.Sprintf("resource.label.namespace_name = %q AND resource.label.pod_name = %s", namespace, podNames[0])
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:		return fmt.Sprintf("resource.label.pod_id = %s", podIDs[0])
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:		return timeSeries.Resource.Labels["pod_id"], nil
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:	return "resource.label.pod_id != \"\" AND resource.label.pod_id != \"machine\""
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:	return fmt.Sprintf("resource.label.namespace_name = %q AND resource.label.pod_name = one_of(%s)", namespace, strings.Join(podNames, ","))
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator.go:	return fmt.Sprintf("resource.label.pod_id = one_of(%s)", strings.Join(podIDs, ","))
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:				"resource.label.pod_id": "my-pod-id",
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			"AND resource.label : pod_name " +
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			"AND resource.label.container_name = \"\" AND resource.label.pod_id != \"\" " +
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			"AND resource.label.pod_id != \"machine\"")
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			"AND resource.label.pod_id = one_of(\"my-pod-id-1\",\"my-pod-id-2\")").
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			"AND resource.label.pod_name = \"my-pod-name\" " +
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			"AND resource.label.pod_name = one_of(\"my-pod-name-1\",\"my-pod-name-2\") " +
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			Resource:   &sd.MonitoredResource{Type: "gke_container", Labels: map[string]string{"pod_id": "my-pod-id"}},
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			Resource:   &sd.MonitoredResource{Type: "gke_container", Labels: map[string]string{"pod_id": "my-pod-id"}},
 custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:			Resource:   &sd.MonitoredResource{Type: "gke_container", Labels: map[string]string{"pod_id": "my-pod-id"}},
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:	req2, _ := labels.NewRequirement("resource.label.pod_name", selection.Exists, []string{})
+custom-metrics-stackdriver-adapter/pkg/adapter/provider/translator_test.go:	req3, _ := labels.NewRequirement("resource.label.pod_name", selection.Exists, []string{})
 kubelet-to-gcm/monitor/controller/translator.go:		"namespace_id":   "",
 kubelet-to-gcm/monitor/controller/translator.go:		"pod_id":         "machine",
 kubelet-to-gcm/monitor/kubelet/translate.go:				"namespace_id":   namespace,

hack/verify-flags/known-flags.txt

@@ -107,6 +107,7 @@ one-time
 on-start
 path-label-config
 pod-id
+pod-name
 pod-scheduled-timeout
 poll-period
 presubmit-jobs
@@ -174,6 +175,7 @@ use-cluster-credentials
 use-ip
 use-kubernetes-cluster-service
 use-new-resource-model
+use-old-resource-model
 use-real-cloud
 use-reviewers
 user-whitelist