fix(endpoints): scrape endpoints and not services

.github/workflows/load_test.yml

@@ -1,9 +1,8 @@
 on:
   push:
-    tags:
-      - v*
     branches:
       - main
+  pull_request:
 
 name: Load Tests
 jobs:

.github/workflows/push_pr.yml

@@ -2,8 +2,6 @@ name: Push/PR pipeline
 
 on:
   push:
-    tags:
-      - v*
     branches:
       - main
   pull_request:

cmd/k8s-target-retriever/main.go

@@ -33,7 +33,7 @@ func main() {
 	}
 
 	kubeconf := endpoints.WithKubeConfig(*kubeConfigFile)
-	ktr, err := endpoints.NewKubernetesTargetRetriever("prometheus.io/scrape", false, kubeconf)
+	ktr, err := endpoints.NewKubernetesTargetRetriever("prometheus.io/scrape", false, true, true, kubeconf)
 	if err != nil {
 		logrus.Fatalf("could not create KubernetesTargetRetriever: %v", err)
 	}
@@ -44,15 +44,16 @@ func main() {
 
 	logrus.Infoln("connected to cluster, watching for targets")
 
-	for range time.Tick(time.Second * 5) {
+	for range time.Tick(time.Second * 7) {
 		targets, err := ktr.GetTargets()
+		logrus.Infof("###################################")
+
 		if err != nil {
 			logrus.Fatalf("could not get targets: %v", err)
 		}
 		for _, b := range targets {
 			logrus.Infof("%s[%s] %s", b.Name, b.Object.Kind, b.URL.String())
 		}
-
-		logrus.Println()
+		logrus.Infof("###################################")
 	}
 }

cmd/nri-prometheus/config.go

@@ -109,6 +109,8 @@ func setViperDefaults(viper *viper.Viper) {
 	viper.SetDefault("insecure_skip_verify", false)
 	viper.SetDefault("standalone", true)
 	viper.SetDefault("disable_autodiscovery", false)
+	viper.SetDefault("scrape_services", true)
+	viper.SetDefault("scrape_endpoints", false)
 	viper.SetDefault("percentiles", []float64{50.0, 95.0, 99.0})
 	viper.SetDefault("worker_threads", 4)
 }

deploy/local.yaml.example

@@ -18,6 +18,7 @@ rules:
     - "nodes/proxy"
     - "pods"
     - "services"
+    - "endpoints"
   verbs: ["get", "list", "watch"]
 - nonResourceURLs:
   - /metrics
@@ -86,7 +87,13 @@ data:
     # standalone: true
     # How often the integration should run. Defaults to 30s.
     # scrape_duration: "30s"
-    # The HTTP client timeout when fetching data from endpoints. Defaults to 30s.
+    # The HTTP client timeout when fetching data from targets. Defaults to 30s.
+    # scrape_services Allows to enable scraping the service and not the endpoints behind.
+    # When endpoints are scraped this is no longer needed
+    scrape_services: true
+    # scrape_endpoints Allows to enable scraping directly endpoints instead of services as prometheus service natively does.
+    # Please notice that depending on the number of endpoints behind a service the load can increase considerably
+    scrape_endpoints: false
     # scrape_timeout: "30s"
     # Wether the integration should run in verbose mode or not. Defaults to false.
     verbose: false

deploy/nri-prometheus.tmpl.yaml

@@ -18,6 +18,7 @@ rules:
     - "nodes/proxy"
     - "pods"
     - "services"
+    - "endpoints"
   verbs: ["get", "list", "watch"]
 - nonResourceURLs:
   - /metrics
@@ -89,7 +90,7 @@ data:
     # How often the integration should run. Defaults to 30s.
     # scrape_duration: "30s"
 
-    # The HTTP client timeout when fetching data from endpoints. Defaults to 5s.
+    # The HTTP client timeout when fetching data from targets. Defaults to 5s.
     # scrape_timeout: "5s"
 
     # How old must the entries used for calculating the counters delta be
@@ -114,11 +115,19 @@ data:
     # The label used to identify scrapable targets. Defaults to "prometheus.io/scrape".
     scrape_enabled_label: "prometheus.io/scrape"
 
+    # scrape_services Allows to enable scraping the service and not the endpoints behind.
+    # When endpoints are scraped this is no longer needed
+    scrape_services: true
+
+    # scrape_endpoints Allows to enable scraping directly endpoints instead of services as prometheus service natively does.
+    # Please notice that depending on the number of endpoints behind a service the load can increase considerably
+    scrape_endpoints: false
+
     # Whether k8s nodes need to be labelled to be scraped or not. Defaults to true.
     require_scrape_enabled_label_for_nodes: true
 
     # Number of worker threads used for scraping targets.
-    # For large clusters with many (>400) endpoints, slowly increase until scrape
+    # For large clusters with many (>400) targets, slowly increase until scrape
     # time falls between the desired `scrape_duration`.
     # Increasing this value too much will result in huge memory consumption if too
     # many metrics are being scraped.

internal/cmd/scraper/scraper.go

@@ -34,6 +34,8 @@ type Config struct {
 	ScrapeTimeout                     time.Duration                `mapstructure:"scrape_timeout"`
 	Standalone                        bool                         `mapstructure:"standalone"`
 	DisableAutodiscovery              bool                         `mapstructure:"disable_autodiscovery"`
+	ScrapeServices                    bool                         `mapstructure:"scrape_services"`
+	ScrapeEndpoints                   bool                         `mapstructure:"scrape_endpoints"`
 	ScrapeDuration                    string                       `mapstructure:"scrape_duration"`
 	EmitterHarvestPeriod              string                       `mapstructure:"emitter_harvest_period"`
 	MinEmitterHarvestPeriod           string                       `mapstructure:"min_emitter_harvest_period"`
@@ -125,7 +127,7 @@ func RunWithEmitters(cfg *Config, emitters []integration.Emitter) error {
 	retrievers = append(retrievers, fixedRetriever)
 
 	if !cfg.DisableAutodiscovery {
-		kubernetesRetriever, err := endpoints.NewKubernetesTargetRetriever(cfg.ScrapeEnabledLabel, cfg.RequireScrapeEnabledLabelForNodes, endpoints.WithInClusterConfig())
+		kubernetesRetriever, err := endpoints.NewKubernetesTargetRetriever(cfg.ScrapeEnabledLabel, cfg.RequireScrapeEnabledLabelForNodes, cfg.ScrapeServices, cfg.ScrapeEndpoints, endpoints.WithInClusterConfig())
 		if err != nil {
 			logrus.WithError(err).Errorf("not possible to get a Kubernetes client. If you aren't running this integration in a Kubernetes cluster, you can ignore this error")
 		} else {

internal/pkg/endpoints/endpoints.go

@@ -18,7 +18,7 @@ type TargetRetriever interface {
 	Name() string
 }
 
-// Object represents a kubernetes object like a pod or a service.
+// Object represents a kubernetes object like a pod or a service or an endpoint.
 type Object struct {
 	Name   string
 	Kind   string