Merge pull request #7 from krancour/exporter-tests

improve test coverage on exporter

exporter/metrics_exporter.go

@@ -14,48 +14,50 @@ import (
 )
 
 type metricsExporter struct {
-	apiClient            sdk.APIClient
+	coreClient           core.APIClient
+	authnClient          authn.APIClient
 	scrapeInterval       time.Duration
-	totalProjects        prometheus.Gauge
-	totalUsers           prometheus.Gauge
-	totalServiceAccounts prometheus.Gauge
+	projectsGauge        prometheus.Gauge
+	usersGauge           prometheus.Gauge
+	serviceAccountsGauge prometheus.Gauge
 	allWorkersByPhase    *prometheus.GaugeVec
-	totalPendingJobs     prometheus.Gauge
+	pendingJobsGauge     prometheus.Gauge
 }
 
 func newMetricsExporter(
 	apiClient sdk.APIClient,
 	scrapeInterval time.Duration,
 ) *metricsExporter {
 	return &metricsExporter{
-		apiClient:      apiClient,
+		coreClient:     apiClient.Core(),
+		authnClient:    apiClient.Authn(),
 		scrapeInterval: scrapeInterval,
-		totalProjects: promauto.NewGauge(
+		projectsGauge: promauto.NewGauge(
 			prometheus.GaugeOpts{
 				Name: "brigade_projects_total",
-				Help: "The total number of brigade projects",
+				Help: "The total number of projects",
 			},
 		),
-		totalUsers: promauto.NewGauge(
+		usersGauge: promauto.NewGauge(
 			prometheus.GaugeOpts{
 				Name: "brigade_users_total",
 				Help: "The total number of users",
 			},
 		),
-		totalServiceAccounts: promauto.NewGauge(
+		serviceAccountsGauge: promauto.NewGauge(
 			prometheus.GaugeOpts{
 				Name: "brigade_service_accounts_total",
 				Help: "The total number of service accounts",
 			},
 		),
 		allWorkersByPhase: promauto.NewGaugeVec(
 			prometheus.GaugeOpts{
-				Name: "brigade_all_workers_by_phase",
-				Help: "All workers separated by phase",
+				Name: "brigade_events_by_worker_phase",
+				Help: "The total number of events grouped by worker phase",
 			},
 			[]string{"workerPhase"},
 		),
-		totalPendingJobs: promauto.NewGauge(
+		pendingJobsGauge: promauto.NewGauge(
 			prometheus.GaugeOpts{
 				Name: "brigade_pending_jobs_total",
 				Help: "The total number of pending jobs",
@@ -70,112 +72,126 @@ func (m *metricsExporter) run(ctx context.Context) {
 	for {
 		select {
 		case <-ticker.C:
-			m.recordMetrics()
+			if err := m.recordProjectsCount(); err != nil {
+				log.Println(err)
+			}
+			if err := m.recordUsersCount(); err != nil {
+				log.Println(err)
+			}
+			if err := m.recordServiceAccountsCount(); err != nil {
+				log.Println(err)
+			}
+			if err := m.recordEventCountsByWorkersPhase(); err != nil {
+				log.Println(err)
+			}
+			if err := m.recordPendingJobsCount(); err != nil {
+				log.Println(err)
+			}
 		case <-ctx.Done():
 			return
 		}
 	}
 }
 
-func (m *metricsExporter) recordMetrics() {
+func (m *metricsExporter) recordProjectsCount() error {
 	// brigade_projects_total
-	projects, err := m.apiClient.Core().Projects().List(
+	projects, err := m.coreClient.Projects().List(
 		context.Background(),
 		&core.ProjectsSelector{},
 		&meta.ListOptions{},
 	)
 	if err != nil {
-		log.Println(err)
-	} else {
-		m.totalProjects.Set(float64(len(projects.Items) +
-			int(projects.RemainingItemCount)))
+		return err
 	}
+	m.projectsGauge.Set(
+		float64(len(projects.Items) + int(projects.RemainingItemCount)),
+	)
+	return nil
+}
 
+func (m *metricsExporter) recordUsersCount() error {
 	// brigade_users_total
-	users, err := m.apiClient.Authn().Users().List(
+	users, err := m.authnClient.Users().List(
 		context.Background(),
 		&authn.UsersSelector{},
 		&meta.ListOptions{},
 	)
 	if err != nil {
-		log.Println(err)
-	} else {
-		m.totalUsers.Set(float64(len(users.Items) +
-			int(users.RemainingItemCount)))
+		return err
 	}
+	m.usersGauge.Set(
+		float64(int64(len(users.Items)) + users.RemainingItemCount),
+	)
+	return nil
+}
 
+func (m *metricsExporter) recordServiceAccountsCount() error {
 	// brigade_service_accounts_total
-	serviceAccounts, err := m.apiClient.Authn().ServiceAccounts().List(
+	serviceAccounts, err := m.authnClient.ServiceAccounts().List(
 		context.Background(),
 		&authn.ServiceAccountsSelector{},
 		&meta.ListOptions{},
 	)
 	if err != nil {
-		log.Println(err)
-	} else {
-		m.totalServiceAccounts.Set(
-			float64(
-				len(serviceAccounts.Items) +
-					int(serviceAccounts.RemainingItemCount),
-			),
-		)
+		return err
 	}
+	m.serviceAccountsGauge.Set(
+		float64(
+			int64(len(serviceAccounts.Items)) + serviceAccounts.RemainingItemCount,
+		),
+	)
+	return nil
+}
 
-	// brigade_all_workers_by_phase
+func (m *metricsExporter) recordEventCountsByWorkersPhase() error {
+	// brigade_events_by_worker_phase
 	for _, phase := range core.WorkerPhasesAll() {
-		var events core.EventList
-		events, err = m.apiClient.Core().Events().List(
+		events, err := m.coreClient.Events().List(
 			context.Background(),
 			&core.EventsSelector{
 				WorkerPhases: []core.WorkerPhase{phase},
 			},
 			&meta.ListOptions{},
 		)
 		if err != nil {
-			log.Println(err)
-		} else {
-			m.allWorkersByPhase.With(
-				prometheus.Labels{"workerPhase": string(phase)},
-			).Set(float64(len(events.Items) + int(events.RemainingItemCount)))
+			return err
 		}
+		m.allWorkersByPhase.With(
+			prometheus.Labels{"workerPhase": string(phase)},
+		).Set(float64(len(events.Items) + int(events.RemainingItemCount)))
+	}
+	return nil
+}
 
-		// brigade_pending_jobs_total
-		//
-		// There is no way to query the API directly for pending Jobs, but only
-		// running Workers should ever HAVE pending Jobs, so if we're currently
-		// counting running Workers, we can iterate over those to count pending
-		// jobs. Note, there's a cap on the max number of workers that can run
-		// concurrently, so we assume that as long as that cap isn't enormous (which
-		// would only occur on an enormous cluster), it's practical to iterate over
-		// all the running workers.
-		if phase == core.WorkerPhaseRunning {
-			var pendingJobs int
-			for {
-				for _, event := range events.Items {
-					for _, job := range event.Worker.Jobs {
-						if job.Status.Phase == core.JobPhasePending {
-							pendingJobs++
-						}
-					}
-				}
-				if events.Continue == "" {
-					break
-				}
-				if events, err = m.apiClient.Core().Events().List(
-					context.Background(),
-					&core.EventsSelector{
-						WorkerPhases: []core.WorkerPhase{phase},
-					},
-					&meta.ListOptions{Continue: events.Continue},
-				); err != nil {
-					log.Println(err)
-					break
+func (m *metricsExporter) recordPendingJobsCount() error {
+	// brigade_pending_jobs_total
+	var pendingJobs int
+	var continueValue string
+	for {
+		events, err := m.coreClient.Events().List(
+			context.Background(),
+			&core.EventsSelector{
+				WorkerPhases: []core.WorkerPhase{core.WorkerPhaseRunning},
+			},
+			&meta.ListOptions{
+				Continue: continueValue,
+			},
+		)
+		if err != nil {
+			return err
+		}
+		for _, event := range events.Items {
+			for _, job := range event.Worker.Jobs {
+				if job.Status.Phase == core.JobPhasePending {
+					pendingJobs++
 				}
 			}
-			if err == nil {
-				m.totalPendingJobs.Set(float64(pendingJobs))
-			}
 		}
+		if events.Continue == "" {
+			break
+		}
+		continueValue = events.Continue
 	}
-
+	m.pendingJobsGauge.Set(float64(pendingJobs))
+	return nil
 }