diff --git a/pkg/services/ngalert/metrics/alertmanager.go b/pkg/services/ngalert/metrics/alertmanager.go
new file mode 100644
index 00000000000..23b85e20723
--- /dev/null
+++ b/pkg/services/ngalert/metrics/alertmanager.go
@@ -0,0 +1,21 @@
+package metrics
+
+import (
+	"fmt"
+
+	"github.com/prometheus/alertmanager/api/metrics"
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+type Alertmanager struct {
+	Registerer prometheus.Registerer
+	*metrics.Alerts
+}
+
+// NewAlertmanagerMetrics creates a set of metrics for the Alertmanager of each organization.
+func NewAlertmanagerMetrics(r prometheus.Registerer) *Alertmanager {
+	return &Alertmanager{
+		Registerer: r,
+		Alerts:     metrics.NewAlerts("grafana", prometheus.WrapRegistererWithPrefix(fmt.Sprintf("%s_%s_", Namespace, Subsystem), r)),
+	}
+}
diff --git a/pkg/services/ngalert/metrics/api.go b/pkg/services/ngalert/metrics/api.go
new file mode 100644
index 00000000000..db97033a0b5
--- /dev/null
+++ b/pkg/services/ngalert/metrics/api.go
@@ -0,0 +1,25 @@
+package metrics
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+type API struct {
+	RequestDuration *prometheus.HistogramVec
+}
+
+func NewAPIMetrics(r prometheus.Registerer) *API {
+	return &API{
+		RequestDuration: promauto.With(r).NewHistogramVec(
+			prometheus.HistogramOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "request_duration_seconds",
+				Help:      "Histogram of requests to the Alerting API",
+				Buckets:   prometheus.DefBuckets,
+			},
+			[]string{"method", "route", "status_code", "backend"},
+		),
+	}
+}
diff --git a/pkg/services/ngalert/metrics/multi_org_alertmanager.go b/pkg/services/ngalert/metrics/multi_org_alertmanager.go
new file mode 100644
index 00000000000..c755457755d
--- /dev/null
+++ b/pkg/services/ngalert/metrics/multi_org_alertmanager.go
@@ -0,0 +1,32 @@
+package metrics
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+type MultiOrgAlertmanager struct {
+	Registerer               prometheus.Registerer
+	ActiveConfigurations     prometheus.Gauge
+	DiscoveredConfigurations prometheus.Gauge
+	registries               *OrgRegistries
+}
+
+func NewMultiOrgAlertmanagerMetrics(r prometheus.Registerer) *MultiOrgAlertmanager {
+	return &MultiOrgAlertmanager{
+		Registerer: r,
+		registries: NewOrgRegistries(),
+		DiscoveredConfigurations: promauto.With(r).NewGauge(prometheus.GaugeOpts{
+			Namespace: Namespace,
+			Subsystem: Subsystem,
+			Name:      "discovered_configurations",
+			Help:      "The number of organizations we've discovered that require an Alertmanager configuration.",
+		}),
+		ActiveConfigurations: promauto.With(r).NewGauge(prometheus.GaugeOpts{
+			Namespace: Namespace,
+			Subsystem: Subsystem,
+			Name:      "active_configurations",
+			Help:      "The number of active Alertmanager configurations.",
+		}),
+	}
+}
diff --git a/pkg/services/ngalert/metrics/ngalert.go b/pkg/services/ngalert/metrics/ngalert.go
index 9681847786a..ae1a304ed34 100644
--- a/pkg/services/ngalert/metrics/ngalert.go
+++ b/pkg/services/ngalert/metrics/ngalert.go
@@ -1,22 +1,7 @@
 package metrics
 
 import (
-	"fmt"
-	"regexp"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/prometheus/alertmanager/api/metrics"
 	"github.com/prometheus/client_golang/prometheus"
-	"github.com/prometheus/client_golang/prometheus/promauto"
-
-	"github.com/grafana/grafana/pkg/api/response"
-	contextmodel "github.com/grafana/grafana/pkg/services/contexthandler/model"
-	apimodels "github.com/grafana/grafana/pkg/services/ngalert/api/tooling/definitions"
-	"github.com/grafana/grafana/pkg/util/ticker"
-
-	"github.com/grafana/grafana/pkg/web"
 )
 
 const (
@@ -37,47 +22,24 @@ func ProvideServiceForTest() *NGAlert {
 }
 
 type NGAlert struct {
-	// Registerer is for use by subcomponents which register their own metrics.
-	Registerer                  prometheus.Registerer
+	// Registerer is used by subcomponents which register their own metrics.
+	Registerer prometheus.Registerer
+
 	schedulerMetrics            *Scheduler
 	stateMetrics                *State
 	multiOrgAlertmanagerMetrics *MultiOrgAlertmanager
 	apiMetrics                  *API
 }
 
-type Scheduler struct {
-	Registerer                          prometheus.Registerer
-	BehindSeconds                       prometheus.Gauge
-	EvalTotal                           *prometheus.CounterVec
-	EvalFailures                        *prometheus.CounterVec
-	EvalDuration                        *prometheus.HistogramVec
-	SchedulePeriodicDuration            prometheus.Histogram
-	SchedulableAlertRules               prometheus.Gauge
-	SchedulableAlertRulesHash           prometheus.Gauge
-	UpdateSchedulableAlertRulesDuration prometheus.Histogram
-	Ticker                              *ticker.Metrics
-	EvaluationMissed                    *prometheus.CounterVec
-}
-
-type MultiOrgAlertmanager struct {
-	Registerer               prometheus.Registerer
-	ActiveConfigurations     prometheus.Gauge
-	DiscoveredConfigurations prometheus.Gauge
-	registries               *OrgRegistries
-}
-
-type API struct {
-	RequestDuration *prometheus.HistogramVec
-}
-
-type Alertmanager struct {
-	Registerer prometheus.Registerer
-	*metrics.Alerts
-}
-
-type State struct {
-	GroupRules *prometheus.GaugeVec
-	AlertState *prometheus.GaugeVec
+// NewNGAlert manages the metrics of all the alerting components.
+func NewNGAlert(r prometheus.Registerer) *NGAlert {
+	return &NGAlert{
+		Registerer:                  r,
+		schedulerMetrics:            NewSchedulerMetrics(r),
+		stateMetrics:                NewStateMetrics(r),
+		multiOrgAlertmanagerMetrics: NewMultiOrgAlertmanagerMetrics(r),
+		apiMetrics:                  NewAPIMetrics(r),
+	}
 }
 
 func (ng *NGAlert) GetSchedulerMetrics() *Scheduler {
@@ -96,25 +58,6 @@ func (ng *NGAlert) GetMultiOrgAlertmanagerMetrics() *MultiOrgAlertmanager {
 	return ng.multiOrgAlertmanagerMetrics
 }
 
-// NewNGAlert manages the metrics of all the alerting components.
-func NewNGAlert(r prometheus.Registerer) *NGAlert {
-	return &NGAlert{
-		Registerer:                  r,
-		schedulerMetrics:            NewSchedulerMetrics(r),
-		stateMetrics:                newStateMetrics(r),
-		multiOrgAlertmanagerMetrics: newMultiOrgAlertmanagerMetrics(r),
-		apiMetrics:                  newAPIMetrics(r),
-	}
-}
-
-// NewAlertmanagerMetrics creates a set of metrics for the Alertmanager of each organization.
-func NewAlertmanagerMetrics(r prometheus.Registerer) *Alertmanager {
-	return &Alertmanager{
-		Registerer: r,
-		Alerts:     metrics.NewAlerts("grafana", prometheus.WrapRegistererWithPrefix(fmt.Sprintf("%s_%s_", Namespace, Subsystem), r)),
-	}
-}
-
 // RemoveOrgRegistry removes the *prometheus.Registry for the specified org. It is safe to call concurrently.
 func (moa *MultiOrgAlertmanager) RemoveOrgRegistry(id int64) {
 	moa.registries.RemoveOrgRegistry(id)
@@ -124,232 +67,3 @@ func (moa *MultiOrgAlertmanager) RemoveOrgRegistry(id int64) {
 func (moa *MultiOrgAlertmanager) GetOrCreateOrgRegistry(id int64) prometheus.Registerer {
 	return moa.registries.GetOrCreateOrgRegistry(id)
 }
-
-func NewSchedulerMetrics(r prometheus.Registerer) *Scheduler {
-	return &Scheduler{
-		Registerer: r,
-		BehindSeconds: promauto.With(r).NewGauge(prometheus.GaugeOpts{
-			Namespace: Namespace,
-			Subsystem: Subsystem,
-			Name:      "scheduler_behind_seconds",
-			Help:      "The total number of seconds the scheduler is behind.",
-		}),
-		// TODO: once rule groups support multiple rules, consider partitioning
-		// on rule group as well as tenant, similar to loki|cortex.
-		EvalTotal: promauto.With(r).NewCounterVec(
-			prometheus.CounterOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "rule_evaluations_total",
-				Help:      "The total number of rule evaluations.",
-			},
-			[]string{"org"},
-		),
-		// TODO: once rule groups support multiple rules, consider partitioning
-		// on rule group as well as tenant, similar to loki|cortex.
-		EvalFailures: promauto.With(r).NewCounterVec(
-			prometheus.CounterOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "rule_evaluation_failures_total",
-				Help:      "The total number of rule evaluation failures.",
-			},
-			[]string{"org"},
-		),
-		EvalDuration: promauto.With(r).NewHistogramVec(
-			prometheus.HistogramOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "rule_evaluation_duration_seconds",
-				Help:      "The duration for a rule to execute.",
-				Buckets:   []float64{.005, .01, .025, .05, .1, .25, .5, 1, 2.5, 5, 10, 25, 50, 100},
-			},
-			[]string{"org"},
-		),
-		SchedulePeriodicDuration: promauto.With(r).NewHistogram(
-			prometheus.HistogramOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "schedule_periodic_duration_seconds",
-				Help:      "The time taken to run the scheduler.",
-				Buckets:   []float64{0.1, 0.25, 0.5, 1, 2, 5, 10},
-			},
-		),
-		SchedulableAlertRules: promauto.With(r).NewGauge(
-			prometheus.GaugeOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "schedule_alert_rules",
-				Help:      "The number of alert rules that could be considered for evaluation at the next tick.",
-			},
-		),
-		SchedulableAlertRulesHash: promauto.With(r).NewGauge(
-			prometheus.GaugeOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "schedule_alert_rules_hash",
-				Help:      "A hash of the alert rules that could be considered for evaluation at the next tick.",
-			}),
-		UpdateSchedulableAlertRulesDuration: promauto.With(r).NewHistogram(
-			prometheus.HistogramOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "schedule_query_alert_rules_duration_seconds",
-				Help:      "The time taken to fetch alert rules from the database.",
-				Buckets:   []float64{0.1, 0.25, 0.5, 1, 2, 5, 10},
-			},
-		),
-		Ticker: ticker.NewMetrics(r, "alerting"),
-		EvaluationMissed: promauto.With(r).NewCounterVec(
-			prometheus.CounterOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "schedule_rule_evaluations_missed_total",
-				Help:      "The total number of rule evaluations missed due to a slow rule evaluation.",
-			},
-			[]string{"org", "name"},
-		),
-	}
-}
-
-func newStateMetrics(r prometheus.Registerer) *State {
-	return &State{
-		// TODO: once rule groups support multiple rules, consider partitioning
-		// on rule group as well as tenant, similar to loki|cortex.
-		GroupRules: promauto.With(r).NewGaugeVec(
-			prometheus.GaugeOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "rule_group_rules",
-				Help:      "The number of rules.",
-			},
-			[]string{"org"},
-		),
-		AlertState: promauto.With(r).NewGaugeVec(prometheus.GaugeOpts{
-			Namespace: Namespace,
-			Subsystem: Subsystem,
-			Name:      "alerts",
-			Help:      "How many alerts by state.",
-		}, []string{"state"}),
-	}
-}
-
-func newMultiOrgAlertmanagerMetrics(r prometheus.Registerer) *MultiOrgAlertmanager {
-	return &MultiOrgAlertmanager{
-		Registerer: r,
-		registries: NewOrgRegistries(),
-		DiscoveredConfigurations: promauto.With(r).NewGauge(prometheus.GaugeOpts{
-			Namespace: Namespace,
-			Subsystem: Subsystem,
-			Name:      "discovered_configurations",
-			Help:      "The number of organizations we've discovered that require an Alertmanager configuration.",
-		}),
-		ActiveConfigurations: promauto.With(r).NewGauge(prometheus.GaugeOpts{
-			Namespace: Namespace,
-			Subsystem: Subsystem,
-			Name:      "active_configurations",
-			Help:      "The number of active Alertmanager configurations.",
-		}),
-	}
-}
-
-func newAPIMetrics(r prometheus.Registerer) *API {
-	return &API{
-		RequestDuration: promauto.With(r).NewHistogramVec(
-			prometheus.HistogramOpts{
-				Namespace: Namespace,
-				Subsystem: Subsystem,
-				Name:      "request_duration_seconds",
-				Help:      "Histogram of requests to the Alerting API",
-				Buckets:   prometheus.DefBuckets,
-			},
-			[]string{"method", "route", "status_code", "backend"},
-		),
-	}
-}
-
-// OrgRegistries represents a map of registries per org.
-type OrgRegistries struct {
-	regsMu sync.Mutex
-	regs   map[int64]prometheus.Registerer
-}
-
-func NewOrgRegistries() *OrgRegistries {
-	return &OrgRegistries{
-		regs: make(map[int64]prometheus.Registerer),
-	}
-}
-
-// GetOrCreateOrgRegistry gets or creates a *prometheus.Registry for the specified org. It is safe to call concurrently.
-func (m *OrgRegistries) GetOrCreateOrgRegistry(orgID int64) prometheus.Registerer {
-	m.regsMu.Lock()
-	defer m.regsMu.Unlock()
-
-	orgRegistry, ok := m.regs[orgID]
-	if !ok {
-		reg := prometheus.NewRegistry()
-		m.regs[orgID] = reg
-		return reg
-	}
-	return orgRegistry
-}
-
-// RemoveOrgRegistry removes the *prometheus.Registry for the specified org. It is safe to call concurrently.
-func (m *OrgRegistries) RemoveOrgRegistry(org int64) {
-	m.regsMu.Lock()
-	defer m.regsMu.Unlock()
-	delete(m.regs, org)
-}
-
-// Instrument wraps a middleware, instrumenting the request latencies.
-func Instrument(
-	method,
-	path string,
-	action func(*contextmodel.ReqContext) response.Response,
-	metrics *API,
-) web.Handler {
-	normalizedPath := MakeLabelValue(path)
-
-	return func(c *contextmodel.ReqContext) {
-		start := time.Now()
-		res := action(c)
-
-		// TODO: We could look up the datasource type via our datasource service
-		var backend string
-		datasourceID := web.Params(c.Req)[":DatasourceID"]
-		if datasourceID == apimodels.GrafanaBackend.String() || datasourceID == "" {
-			backend = GrafanaBackend
-		} else {
-			backend = ProxyBackend
-		}
-
-		ls := prometheus.Labels{
-			"method":      method,
-			"route":       normalizedPath,
-			"status_code": fmt.Sprint(res.Status()),
-			"backend":     backend,
-		}
-		res.WriteTo(c)
-		metrics.RequestDuration.With(ls).Observe(time.Since(start).Seconds())
-	}
-}
-
-var invalidChars = regexp.MustCompile(`[^a-zA-Z0-9]+`)
-
-// MakeLabelValue normalizes a path template
-func MakeLabelValue(path string) string {
-	// Convert non-alnums to underscores.
-	result := invalidChars.ReplaceAllString(path, "_")
-
-	// Trim leading and trailing underscores.
-	result = strings.Trim(result, "_")
-
-	// Make it all lowercase
-	result = strings.ToLower(result)
-
-	// Special case.
-	if result == "" {
-		result = "root"
-	}
-	return result
-}
diff --git a/pkg/services/ngalert/metrics/scheduler.go b/pkg/services/ngalert/metrics/scheduler.go
new file mode 100644
index 00000000000..fb1523918a3
--- /dev/null
+++ b/pkg/services/ngalert/metrics/scheduler.go
@@ -0,0 +1,108 @@
+package metrics
+
+import (
+	"github.com/grafana/grafana/pkg/util/ticker"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+type Scheduler struct {
+	Registerer                          prometheus.Registerer
+	BehindSeconds                       prometheus.Gauge
+	EvalTotal                           *prometheus.CounterVec
+	EvalFailures                        *prometheus.CounterVec
+	EvalDuration                        *prometheus.HistogramVec
+	SchedulePeriodicDuration            prometheus.Histogram
+	SchedulableAlertRules               prometheus.Gauge
+	SchedulableAlertRulesHash           prometheus.Gauge
+	UpdateSchedulableAlertRulesDuration prometheus.Histogram
+	Ticker                              *ticker.Metrics
+	EvaluationMissed                    *prometheus.CounterVec
+}
+
+func NewSchedulerMetrics(r prometheus.Registerer) *Scheduler {
+	return &Scheduler{
+		Registerer: r,
+		BehindSeconds: promauto.With(r).NewGauge(prometheus.GaugeOpts{
+			Namespace: Namespace,
+			Subsystem: Subsystem,
+			Name:      "scheduler_behind_seconds",
+			Help:      "The total number of seconds the scheduler is behind.",
+		}),
+		// TODO: once rule groups support multiple rules, consider partitioning
+		// on rule group as well as tenant, similar to loki|cortex.
+		EvalTotal: promauto.With(r).NewCounterVec(
+			prometheus.CounterOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "rule_evaluations_total",
+				Help:      "The total number of rule evaluations.",
+			},
+			[]string{"org"},
+		),
+		// TODO: once rule groups support multiple rules, consider partitioning
+		// on rule group as well as tenant, similar to loki|cortex.
+		EvalFailures: promauto.With(r).NewCounterVec(
+			prometheus.CounterOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "rule_evaluation_failures_total",
+				Help:      "The total number of rule evaluation failures.",
+			},
+			[]string{"org"},
+		),
+		EvalDuration: promauto.With(r).NewHistogramVec(
+			prometheus.HistogramOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "rule_evaluation_duration_seconds",
+				Help:      "The duration for a rule to execute.",
+				Buckets:   []float64{.005, .01, .025, .05, .1, .25, .5, 1, 2.5, 5, 10, 25, 50, 100},
+			},
+			[]string{"org"},
+		),
+		SchedulePeriodicDuration: promauto.With(r).NewHistogram(
+			prometheus.HistogramOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "schedule_periodic_duration_seconds",
+				Help:      "The time taken to run the scheduler.",
+				Buckets:   []float64{0.1, 0.25, 0.5, 1, 2, 5, 10},
+			},
+		),
+		SchedulableAlertRules: promauto.With(r).NewGauge(
+			prometheus.GaugeOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "schedule_alert_rules",
+				Help:      "The number of alert rules that could be considered for evaluation at the next tick.",
+			},
+		),
+		SchedulableAlertRulesHash: promauto.With(r).NewGauge(
+			prometheus.GaugeOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "schedule_alert_rules_hash",
+				Help:      "A hash of the alert rules that could be considered for evaluation at the next tick.",
+			}),
+		UpdateSchedulableAlertRulesDuration: promauto.With(r).NewHistogram(
+			prometheus.HistogramOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "schedule_query_alert_rules_duration_seconds",
+				Help:      "The time taken to fetch alert rules from the database.",
+				Buckets:   []float64{0.1, 0.25, 0.5, 1, 2, 5, 10},
+			},
+		),
+		Ticker: ticker.NewMetrics(r, "alerting"),
+		EvaluationMissed: promauto.With(r).NewCounterVec(
+			prometheus.CounterOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "schedule_rule_evaluations_missed_total",
+				Help:      "The total number of rule evaluations missed due to a slow rule evaluation.",
+			},
+			[]string{"org", "name"},
+		),
+	}
+}
diff --git a/pkg/services/ngalert/metrics/state.go b/pkg/services/ngalert/metrics/state.go
new file mode 100644
index 00000000000..2a5bf072b1a
--- /dev/null
+++ b/pkg/services/ngalert/metrics/state.go
@@ -0,0 +1,33 @@
+package metrics
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+type State struct {
+	GroupRules *prometheus.GaugeVec
+	AlertState *prometheus.GaugeVec
+}
+
+func NewStateMetrics(r prometheus.Registerer) *State {
+	return &State{
+		// TODO: once rule groups support multiple rules, consider partitioning
+		// on rule group as well as tenant, similar to loki|cortex.
+		GroupRules: promauto.With(r).NewGaugeVec(
+			prometheus.GaugeOpts{
+				Namespace: Namespace,
+				Subsystem: Subsystem,
+				Name:      "rule_group_rules",
+				Help:      "The number of rules.",
+			},
+			[]string{"org"},
+		),
+		AlertState: promauto.With(r).NewGaugeVec(prometheus.GaugeOpts{
+			Namespace: Namespace,
+			Subsystem: Subsystem,
+			Name:      "alerts",
+			Help:      "How many alerts by state.",
+		}, []string{"state"}),
+	}
+}
diff --git a/pkg/services/ngalert/metrics/util.go b/pkg/services/ngalert/metrics/util.go
new file mode 100644
index 00000000000..db65e1e2b87
--- /dev/null
+++ b/pkg/services/ngalert/metrics/util.go
@@ -0,0 +1,104 @@
+package metrics
+
+import (
+	"fmt"
+	"regexp"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/grafana/grafana/pkg/web"
+
+	"github.com/grafana/grafana/pkg/api/response"
+
+	"github.com/prometheus/client_golang/prometheus"
+
+	contextmodel "github.com/grafana/grafana/pkg/services/contexthandler/model"
+	apimodels "github.com/grafana/grafana/pkg/services/ngalert/api/tooling/definitions"
+)
+
+// OrgRegistries represents a map of registries per org.
+type OrgRegistries struct {
+	regsMu sync.Mutex
+	regs   map[int64]prometheus.Registerer
+}
+
+func NewOrgRegistries() *OrgRegistries {
+	return &OrgRegistries{
+		regs: make(map[int64]prometheus.Registerer),
+	}
+}
+
+// GetOrCreateOrgRegistry gets or creates a *prometheus.Registry for the specified org. It is safe to call concurrently.
+func (m *OrgRegistries) GetOrCreateOrgRegistry(orgID int64) prometheus.Registerer {
+	m.regsMu.Lock()
+	defer m.regsMu.Unlock()
+
+	orgRegistry, ok := m.regs[orgID]
+	if !ok {
+		reg := prometheus.NewRegistry()
+		m.regs[orgID] = reg
+		return reg
+	}
+	return orgRegistry
+}
+
+// RemoveOrgRegistry removes the *prometheus.Registry for the specified org. It is safe to call concurrently.
+func (m *OrgRegistries) RemoveOrgRegistry(org int64) {
+	m.regsMu.Lock()
+	defer m.regsMu.Unlock()
+	delete(m.regs, org)
+}
+
+// Instrument wraps a middleware, instrumenting the request latencies.
+func Instrument(
+	method,
+	path string,
+	action func(*contextmodel.ReqContext) response.Response,
+	metrics *API,
+) web.Handler {
+	normalizedPath := MakeLabelValue(path)
+
+	return func(c *contextmodel.ReqContext) {
+		start := time.Now()
+		res := action(c)
+
+		// TODO: We could look up the datasource type via our datasource service
+		var backend string
+		datasourceID := web.Params(c.Req)[":DatasourceID"]
+		if datasourceID == apimodels.GrafanaBackend.String() || datasourceID == "" {
+			backend = GrafanaBackend
+		} else {
+			backend = ProxyBackend
+		}
+
+		ls := prometheus.Labels{
+			"method":      method,
+			"route":       normalizedPath,
+			"status_code": fmt.Sprint(res.Status()),
+			"backend":     backend,
+		}
+		res.WriteTo(c)
+		metrics.RequestDuration.With(ls).Observe(time.Since(start).Seconds())
+	}
+}
+
+var invalidChars = regexp.MustCompile(`[^a-zA-Z0-9]+`)
+
+// MakeLabelValue normalizes a path template
+func MakeLabelValue(path string) string {
+	// Convert non-alnums to underscores.
+	result := invalidChars.ReplaceAllString(path, "_")
+
+	// Trim leading and trailing underscores.
+	result = strings.Trim(result, "_")
+
+	// Make it all lowercase
+	result = strings.ToLower(result)
+
+	// Special case.
+	if result == "" {
+		result = "root"
+	}
+	return result
+}