feat: Add Metrics V3 API (#19068)

Metrics v3 is mainly a reorganization of metrics into smaller groups of
metrics and the removal of internal aggregation of metrics received from
peer nodes in a MinIO cluster.

This change adds the endpoint `/minio/metrics/v3` as the top-level metrics
endpoint and under this, various sub-endpoints are implemented. These
are currently documented in `docs/metrics/v3.md`

The handler will serve metrics at any path
`/minio/metrics/v3/PATH`, as follows:

when PATH is a sub-endpoint listed above => serves the group of
metrics under that path; or when PATH is a (non-empty) parent 
directory of the sub-endpoints listed above => serves metrics
from each child sub-endpoint of PATH. otherwise, returns a no 
resource found error

All available metrics are listed in the `docs/metrics/v3.md`. More will
be added subsequently.

cmd/data-usage-utils.go

@@ -140,7 +140,7 @@ func (dui DataUsageInfo) tierStats() []madmin.TierInfo {
 	return infos
 }
 
-func (dui DataUsageInfo) tierMetrics() (metrics []Metric) {
+func (dui DataUsageInfo) tierMetrics() (metrics []MetricV2) {
 	if dui.TierStats == nil {
 		return nil
 	}
@@ -148,17 +148,17 @@ func (dui DataUsageInfo) tierMetrics() (metrics []Metric) {
 	//     minio_cluster_ilm_transitioned_objects{tier="S3TIER-1"}=1
 	//     minio_cluster_ilm_transitioned_versions{tier="S3TIER-1"}=3
 	for tier, st := range dui.TierStats.Tiers {
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    getClusterTransitionedBytesMD(),
 			Value:          float64(st.TotalSize),
 			VariableLabels: map[string]string{"tier": tier},
 		})
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    getClusterTransitionedObjectsMD(),
 			Value:          float64(st.NumObjects),
 			VariableLabels: map[string]string{"tier": tier},
 		})
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    getClusterTransitionedVersionsMD(),
 			Value:          float64(st.NumVersions),
 			VariableLabels: map[string]string{"tier": tier},

cmd/generic-handlers.go

@@ -233,7 +233,8 @@ func guessIsMetricsReq(req *http.Request) bool {
 		req.URL.Path == minioReservedBucketPath+prometheusMetricsV2ClusterPath ||
 		req.URL.Path == minioReservedBucketPath+prometheusMetricsV2NodePath ||
 		req.URL.Path == minioReservedBucketPath+prometheusMetricsV2BucketPath ||
-		req.URL.Path == minioReservedBucketPath+prometheusMetricsV2ResourcePath
+		req.URL.Path == minioReservedBucketPath+prometheusMetricsV2ResourcePath ||
+		strings.HasPrefix(req.URL.Path, minioReservedBucketPath+metricsV3Path)
 }
 
 // guessIsRPCReq - returns true if the request is for an RPC endpoint.

cmd/globals.go

@@ -470,6 +470,7 @@ var (
 
 	// Indicates if server was started as `--address ":0"`
 	globalDynamicAPIPort bool
+
 	// Add new variable global values here.
 )
 

cmd/http-stats.go

@@ -269,6 +269,28 @@ func (s *bucketConnStats) getS3InOutBytes() map[string]inOutBytes {
 	return bucketStats
 }
 
+// Return S3 total input/output bytes for each
+func (s *bucketConnStats) getBucketS3InOutBytes(buckets []string) map[string]inOutBytes {
+	s.RLock()
+	defer s.RUnlock()
+
+	if len(s.stats) == 0 || len(buckets) == 0 {
+		return nil
+	}
+
+	bucketStats := make(map[string]inOutBytes, len(buckets))
+	for _, bucket := range buckets {
+		if stats, ok := s.stats[bucket]; ok {
+			bucketStats[bucket] = inOutBytes{
+				In:  stats.s3InputBytes,
+				Out: stats.s3OutputBytes,
+			}
+		}
+	}
+
+	return bucketStats
+}
+
 // delete metrics once bucket is deleted.
 func (s *bucketConnStats) delete(bucket string) {
 	s.Lock()

cmd/metrics-resource.go

@@ -81,7 +81,7 @@ var (
 	resourceMetricsMapMu sync.RWMutex
 	// resourceMetricsHelpMap maps metric name to its help string
 	resourceMetricsHelpMap map[MetricName]string
-	resourceMetricsGroups  []*MetricsGroup
+	resourceMetricsGroups  []*MetricsGroupV2
 	// initial values for drives (at the time  of server startup)
 	// used for calculating avg values for drive metrics
 	latestDriveStats      map[string]madmin.DiskIOStats
@@ -164,7 +164,7 @@ func init() {
 		cpuLoad5Perc:      "CPU load average 5min (percentage)",
 		cpuLoad15Perc:     "CPU load average 15min (percentage)",
 	}
-	resourceMetricsGroups = []*MetricsGroup{
+	resourceMetricsGroups = []*MetricsGroupV2{
 		getResourceMetrics(),
 	}
 
@@ -405,7 +405,7 @@ func startResourceMetricsCollection() {
 
 // minioResourceCollector is the Collector for resource metrics
 type minioResourceCollector struct {
-	metricsGroups []*MetricsGroup
+	metricsGroups []*MetricsGroupV2
 	desc          *prometheus.Desc
 }
 
@@ -417,7 +417,7 @@ func (c *minioResourceCollector) Describe(ch chan<- *prometheus.Desc) {
 // Collect is called by the Prometheus registry when collecting metrics.
 func (c *minioResourceCollector) Collect(out chan<- prometheus.Metric) {
 	var wg sync.WaitGroup
-	publish := func(in <-chan Metric) {
+	publish := func(in <-chan MetricV2) {
 		defer wg.Done()
 		for metric := range in {
 			labels, values := getOrderedLabelValueArrays(metric.VariableLabels)
@@ -436,18 +436,18 @@ func (c *minioResourceCollector) Collect(out chan<- prometheus.Metric) {
 // and returns reference of minio resource Collector
 // It creates the Prometheus Description which is used
 // to define Metric and  help string
-func newMinioResourceCollector(metricsGroups []*MetricsGroup) *minioResourceCollector {
+func newMinioResourceCollector(metricsGroups []*MetricsGroupV2) *minioResourceCollector {
 	return &minioResourceCollector{
 		metricsGroups: metricsGroups,
 		desc:          prometheus.NewDesc("minio_resource_stats", "Resource statistics exposed by MinIO server", nil, nil),
 	}
 }
 
-func prepareResourceMetrics(rm ResourceMetric, subSys MetricSubsystem, requireAvgMax bool) []Metric {
+func prepareResourceMetrics(rm ResourceMetric, subSys MetricSubsystem, requireAvgMax bool) []MetricV2 {
 	help := resourceMetricsHelpMap[rm.Name]
 	name := rm.Name
-	metrics := make([]Metric, 0, 3)
-	metrics = append(metrics, Metric{
+	metrics := make([]MetricV2, 0, 3)
+	metrics = append(metrics, MetricV2{
 		Description:    getResourceMetricDescription(subSys, name, help),
 		Value:          rm.Current,
 		VariableLabels: cloneMSS(rm.Labels),
@@ -456,7 +456,7 @@ func prepareResourceMetrics(rm ResourceMetric, subSys MetricSubsystem, requireAv
 	if requireAvgMax {
 		avgName := MetricName(fmt.Sprintf("%s_avg", name))
 		avgHelp := fmt.Sprintf("%s (avg)", help)
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    getResourceMetricDescription(subSys, avgName, avgHelp),
 			Value:          math.Round(rm.Avg*100) / 100,
 			VariableLabels: cloneMSS(rm.Labels),
@@ -464,7 +464,7 @@ func prepareResourceMetrics(rm ResourceMetric, subSys MetricSubsystem, requireAv
 
 		maxName := MetricName(fmt.Sprintf("%s_max", name))
 		maxHelp := fmt.Sprintf("%s (max)", help)
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    getResourceMetricDescription(subSys, maxName, maxHelp),
 			Value:          rm.Max,
 			VariableLabels: cloneMSS(rm.Labels),
@@ -484,12 +484,12 @@ func getResourceMetricDescription(subSys MetricSubsystem, name MetricName, help
 	}
 }
 
-func getResourceMetrics() *MetricsGroup {
-	mg := &MetricsGroup{
+func getResourceMetrics() *MetricsGroupV2 {
+	mg := &MetricsGroupV2{
 		cacheInterval: resourceMetricsCacheInterval,
 	}
-	mg.RegisterRead(func(ctx context.Context) []Metric {
-		metrics := []Metric{}
+	mg.RegisterRead(func(ctx context.Context) []MetricV2 {
+		metrics := []MetricV2{}
 
 		subSystems := []MetricSubsystem{interfaceSubsystem, memSubsystem, driveSubsystem, cpuSubsystem}
 		resourceMetricsMapMu.RLock()

cmd/metrics-router.go

@@ -18,6 +18,7 @@
 package cmd
 
 import (
+	"net/http"
 	"strings"
 
 	"github.com/minio/mux"
@@ -30,6 +31,9 @@ const (
 	prometheusMetricsV2BucketPath   = "/v2/metrics/bucket"
 	prometheusMetricsV2NodePath     = "/v2/metrics/node"
 	prometheusMetricsV2ResourcePath = "/v2/metrics/resource"
+
+	// Metrics v3 endpoints
+	metricsV3Path = "/metrics/v3"
 )
 
 // Standard env prometheus auth type
@@ -48,10 +52,10 @@ const (
 func registerMetricsRouter(router *mux.Router) {
 	// metrics router
 	metricsRouter := router.NewRoute().PathPrefix(minioReservedBucketPath).Subrouter()
-	authType := strings.ToLower(env.Get(EnvPrometheusAuthType, string(prometheusJWT)))
+	authType := prometheusAuthType(strings.ToLower(env.Get(EnvPrometheusAuthType, string(prometheusJWT))))
 
 	auth := AuthMiddleware
-	if prometheusAuthType(authType) == prometheusPublic {
+	if authType == prometheusPublic {
 		auth = NoAuthMiddleware
 	}
 	metricsRouter.Handle(prometheusMetricsPathLegacy, auth(metricsHandler()))
@@ -59,4 +63,11 @@ func registerMetricsRouter(router *mux.Router) {
 	metricsRouter.Handle(prometheusMetricsV2BucketPath, auth(metricsBucketHandler()))
 	metricsRouter.Handle(prometheusMetricsV2NodePath, auth(metricsNodeHandler()))
 	metricsRouter.Handle(prometheusMetricsV2ResourcePath, auth(metricsResourceHandler()))
+
+	// Metrics v3!
+	metricsV3Server := newMetricsV3Server(authType)
+
+	// Register metrics v3 handler. It also accepts an optional query
+	// parameter `?list` - see handler for details.
+	metricsRouter.Methods(http.MethodGet).Path(metricsV3Path + "{pathComps:.*}").Handler(metricsV3Server)
 }

cmd/metrics-v2_gen.go

@@ -7,7 +7,223 @@ import (
 )
 
 // MarshalMsg implements msgp.Marshaler
-func (z *Metric) MarshalMsg(b []byte) (o []byte, err error) {
+func (z *MetricDescription) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	// map header, size 5
+	// string "Namespace"
+	o = append(o, 0x85, 0xa9, 0x4e, 0x61, 0x6d, 0x65, 0x73, 0x70, 0x61, 0x63, 0x65)
+	o = msgp.AppendString(o, string(z.Namespace))
+	// string "Subsystem"
+	o = append(o, 0xa9, 0x53, 0x75, 0x62, 0x73, 0x79, 0x73, 0x74, 0x65, 0x6d)
+	o = msgp.AppendString(o, string(z.Subsystem))
+	// string "Name"
+	o = append(o, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
+	o = msgp.AppendString(o, string(z.Name))
+	// string "Help"
+	o = append(o, 0xa4, 0x48, 0x65, 0x6c, 0x70)
+	o = msgp.AppendString(o, z.Help)
+	// string "Type"
+	o = append(o, 0xa4, 0x54, 0x79, 0x70, 0x65)
+	o = msgp.AppendString(o, string(z.Type))
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *MetricDescription) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	var field []byte
+	_ = field
+	var zb0001 uint32
+	zb0001, bts, err = msgp.ReadMapHeaderBytes(bts)
+	if err != nil {
+		err = msgp.WrapError(err)
+		return
+	}
+	for zb0001 > 0 {
+		zb0001--
+		field, bts, err = msgp.ReadMapKeyZC(bts)
+		if err != nil {
+			err = msgp.WrapError(err)
+			return
+		}
+		switch msgp.UnsafeString(field) {
+		case "Namespace":
+			{
+				var zb0002 string
+				zb0002, bts, err = msgp.ReadStringBytes(bts)
+				if err != nil {
+					err = msgp.WrapError(err, "Namespace")
+					return
+				}
+				z.Namespace = MetricNamespace(zb0002)
+			}
+		case "Subsystem":
+			{
+				var zb0003 string
+				zb0003, bts, err = msgp.ReadStringBytes(bts)
+				if err != nil {
+					err = msgp.WrapError(err, "Subsystem")
+					return
+				}
+				z.Subsystem = MetricSubsystem(zb0003)
+			}
+		case "Name":
+			{
+				var zb0004 string
+				zb0004, bts, err = msgp.ReadStringBytes(bts)
+				if err != nil {
+					err = msgp.WrapError(err, "Name")
+					return
+				}
+				z.Name = MetricName(zb0004)
+			}
+		case "Help":
+			z.Help, bts, err = msgp.ReadStringBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "Help")
+				return
+			}
+		case "Type":
+			{
+				var zb0005 string
+				zb0005, bts, err = msgp.ReadStringBytes(bts)
+				if err != nil {
+					err = msgp.WrapError(err, "Type")
+					return
+				}
+				z.Type = MetricTypeV2(zb0005)
+			}
+		default:
+			bts, err = msgp.Skip(bts)
+			if err != nil {
+				err = msgp.WrapError(err)
+				return
+			}
+		}
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z *MetricDescription) Msgsize() (s int) {
+	s = 1 + 10 + msgp.StringPrefixSize + len(string(z.Namespace)) + 10 + msgp.StringPrefixSize + len(string(z.Subsystem)) + 5 + msgp.StringPrefixSize + len(string(z.Name)) + 5 + msgp.StringPrefixSize + len(z.Help) + 5 + msgp.StringPrefixSize + len(string(z.Type))
+	return
+}
+
+// MarshalMsg implements msgp.Marshaler
+func (z MetricName) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	o = msgp.AppendString(o, string(z))
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *MetricName) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	{
+		var zb0001 string
+		zb0001, bts, err = msgp.ReadStringBytes(bts)
+		if err != nil {
+			err = msgp.WrapError(err)
+			return
+		}
+		(*z) = MetricName(zb0001)
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z MetricName) Msgsize() (s int) {
+	s = msgp.StringPrefixSize + len(string(z))
+	return
+}
+
+// MarshalMsg implements msgp.Marshaler
+func (z MetricNamespace) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	o = msgp.AppendString(o, string(z))
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *MetricNamespace) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	{
+		var zb0001 string
+		zb0001, bts, err = msgp.ReadStringBytes(bts)
+		if err != nil {
+			err = msgp.WrapError(err)
+			return
+		}
+		(*z) = MetricNamespace(zb0001)
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z MetricNamespace) Msgsize() (s int) {
+	s = msgp.StringPrefixSize + len(string(z))
+	return
+}
+
+// MarshalMsg implements msgp.Marshaler
+func (z MetricSubsystem) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	o = msgp.AppendString(o, string(z))
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *MetricSubsystem) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	{
+		var zb0001 string
+		zb0001, bts, err = msgp.ReadStringBytes(bts)
+		if err != nil {
+			err = msgp.WrapError(err)
+			return
+		}
+		(*z) = MetricSubsystem(zb0001)
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z MetricSubsystem) Msgsize() (s int) {
+	s = msgp.StringPrefixSize + len(string(z))
+	return
+}
+
+// MarshalMsg implements msgp.Marshaler
+func (z MetricTypeV2) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	o = msgp.AppendString(o, string(z))
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *MetricTypeV2) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	{
+		var zb0001 string
+		zb0001, bts, err = msgp.ReadStringBytes(bts)
+		if err != nil {
+			err = msgp.WrapError(err)
+			return
+		}
+		(*z) = MetricTypeV2(zb0001)
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z MetricTypeV2) Msgsize() (s int) {
+	s = msgp.StringPrefixSize + len(string(z))
+	return
+}
+
+// MarshalMsg implements msgp.Marshaler
+func (z *MetricV2) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
 	// map header, size 6
 	// string "Description"
@@ -48,7 +264,7 @@ func (z *Metric) MarshalMsg(b []byte) (o []byte, err error) {
 }
 
 // UnmarshalMsg implements msgp.Unmarshaler
-func (z *Metric) UnmarshalMsg(bts []byte) (o []byte, err error) {
+func (z *MetricV2) UnmarshalMsg(bts []byte) (o []byte, err error) {
 	var field []byte
 	_ = field
 	var zb0001 uint32
@@ -186,7 +402,7 @@ func (z *Metric) UnmarshalMsg(bts []byte) (o []byte, err error) {
 }
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z *Metric) Msgsize() (s int) {
+func (z *MetricV2) Msgsize() (s int) {
 	s = 1 + 12 + z.Description.Msgsize() + 13 + msgp.MapHeaderSize
 	if z.StaticLabels != nil {
 		for za0001, za0002 := range z.StaticLabels {
@@ -211,287 +427,6 @@ func (z *Metric) Msgsize() (s int) {
 	return
 }
 
-// MarshalMsg implements msgp.Marshaler
-func (z *MetricDescription) MarshalMsg(b []byte) (o []byte, err error) {
-	o = msgp.Require(b, z.Msgsize())
-	// map header, size 5
-	// string "Namespace"
-	o = append(o, 0x85, 0xa9, 0x4e, 0x61, 0x6d, 0x65, 0x73, 0x70, 0x61, 0x63, 0x65)
-	o = msgp.AppendString(o, string(z.Namespace))
-	// string "Subsystem"
-	o = append(o, 0xa9, 0x53, 0x75, 0x62, 0x73, 0x79, 0x73, 0x74, 0x65, 0x6d)
-	o = msgp.AppendString(o, string(z.Subsystem))
-	// string "Name"
-	o = append(o, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
-	o = msgp.AppendString(o, string(z.Name))
-	// string "Help"
-	o = append(o, 0xa4, 0x48, 0x65, 0x6c, 0x70)
-	o = msgp.AppendString(o, z.Help)
-	// string "Type"
-	o = append(o, 0xa4, 0x54, 0x79, 0x70, 0x65)
-	o = msgp.AppendString(o, string(z.Type))
-	return
-}
-
-// UnmarshalMsg implements msgp.Unmarshaler
-func (z *MetricDescription) UnmarshalMsg(bts []byte) (o []byte, err error) {
-	var field []byte
-	_ = field
-	var zb0001 uint32
-	zb0001, bts, err = msgp.ReadMapHeaderBytes(bts)
-	if err != nil {
-		err = msgp.WrapError(err)
-		return
-	}
-	for zb0001 > 0 {
-		zb0001--
-		field, bts, err = msgp.ReadMapKeyZC(bts)
-		if err != nil {
-			err = msgp.WrapError(err)
-			return
-		}
-		switch msgp.UnsafeString(field) {
-		case "Namespace":
-			{
-				var zb0002 string
-				zb0002, bts, err = msgp.ReadStringBytes(bts)
-				if err != nil {
-					err = msgp.WrapError(err, "Namespace")
-					return
-				}
-				z.Namespace = MetricNamespace(zb0002)
-			}
-		case "Subsystem":
-			{
-				var zb0003 string
-				zb0003, bts, err = msgp.ReadStringBytes(bts)
-				if err != nil {
-					err = msgp.WrapError(err, "Subsystem")
-					return
-				}
-				z.Subsystem = MetricSubsystem(zb0003)
-			}
-		case "Name":
-			{
-				var zb0004 string
-				zb0004, bts, err = msgp.ReadStringBytes(bts)
-				if err != nil {
-					err = msgp.WrapError(err, "Name")
-					return
-				}
-				z.Name = MetricName(zb0004)
-			}
-		case "Help":
-			z.Help, bts, err = msgp.ReadStringBytes(bts)
-			if err != nil {
-				err = msgp.WrapError(err, "Help")
-				return
-			}
-		case "Type":
-			{
-				var zb0005 string
-				zb0005, bts, err = msgp.ReadStringBytes(bts)
-				if err != nil {
-					err = msgp.WrapError(err, "Type")
-					return
-				}
-				z.Type = MetricType(zb0005)
-			}
-		default:
-			bts, err = msgp.Skip(bts)
-			if err != nil {
-				err = msgp.WrapError(err)
-				return
-			}
-		}
-	}
-	o = bts
-	return
-}
-
-// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z *MetricDescription) Msgsize() (s int) {
-	s = 1 + 10 + msgp.StringPrefixSize + len(string(z.Namespace)) + 10 + msgp.StringPrefixSize + len(string(z.Subsystem)) + 5 + msgp.StringPrefixSize + len(string(z.Name)) + 5 + msgp.StringPrefixSize + len(z.Help) + 5 + msgp.StringPrefixSize + len(string(z.Type))
-	return
-}
-
-// MarshalMsg implements msgp.Marshaler
-func (z MetricName) MarshalMsg(b []byte) (o []byte, err error) {
-	o = msgp.Require(b, z.Msgsize())
-	o = msgp.AppendString(o, string(z))
-	return
-}
-
-// UnmarshalMsg implements msgp.Unmarshaler
-func (z *MetricName) UnmarshalMsg(bts []byte) (o []byte, err error) {
-	{
-		var zb0001 string
-		zb0001, bts, err = msgp.ReadStringBytes(bts)
-		if err != nil {
-			err = msgp.WrapError(err)
-			return
-		}
-		(*z) = MetricName(zb0001)
-	}
-	o = bts
-	return
-}
-
-// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z MetricName) Msgsize() (s int) {
-	s = msgp.StringPrefixSize + len(string(z))
-	return
-}
-
-// MarshalMsg implements msgp.Marshaler
-func (z MetricNamespace) MarshalMsg(b []byte) (o []byte, err error) {
-	o = msgp.Require(b, z.Msgsize())
-	o = msgp.AppendString(o, string(z))
-	return
-}
-
-// UnmarshalMsg implements msgp.Unmarshaler
-func (z *MetricNamespace) UnmarshalMsg(bts []byte) (o []byte, err error) {
-	{
-		var zb0001 string
-		zb0001, bts, err = msgp.ReadStringBytes(bts)
-		if err != nil {
-			err = msgp.WrapError(err)
-			return
-		}
-		(*z) = MetricNamespace(zb0001)
-	}
-	o = bts
-	return
-}
-
-// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z MetricNamespace) Msgsize() (s int) {
-	s = msgp.StringPrefixSize + len(string(z))
-	return
-}
-
-// MarshalMsg implements msgp.Marshaler
-func (z MetricSubsystem) MarshalMsg(b []byte) (o []byte, err error) {
-	o = msgp.Require(b, z.Msgsize())
-	o = msgp.AppendString(o, string(z))
-	return
-}
-
-// UnmarshalMsg implements msgp.Unmarshaler
-func (z *MetricSubsystem) UnmarshalMsg(bts []byte) (o []byte, err error) {
-	{
-		var zb0001 string
-		zb0001, bts, err = msgp.ReadStringBytes(bts)
-		if err != nil {
-			err = msgp.WrapError(err)
-			return
-		}
-		(*z) = MetricSubsystem(zb0001)
-	}
-	o = bts
-	return
-}
-
-// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z MetricSubsystem) Msgsize() (s int) {
-	s = msgp.StringPrefixSize + len(string(z))
-	return
-}
-
-// MarshalMsg implements msgp.Marshaler
-func (z MetricType) MarshalMsg(b []byte) (o []byte, err error) {
-	o = msgp.Require(b, z.Msgsize())
-	o = msgp.AppendString(o, string(z))
-	return
-}
-
-// UnmarshalMsg implements msgp.Unmarshaler
-func (z *MetricType) UnmarshalMsg(bts []byte) (o []byte, err error) {
-	{
-		var zb0001 string
-		zb0001, bts, err = msgp.ReadStringBytes(bts)
-		if err != nil {
-			err = msgp.WrapError(err)
-			return
-		}
-		(*z) = MetricType(zb0001)
-	}
-	o = bts
-	return
-}
-
-// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z MetricType) Msgsize() (s int) {
-	s = msgp.StringPrefixSize + len(string(z))
-	return
-}
-
-// MarshalMsg implements msgp.Marshaler
-func (z *MetricsGroup) MarshalMsg(b []byte) (o []byte, err error) {
-	o = msgp.Require(b, z.Msgsize())
-	// map header, size 2
-	// string "cacheInterval"
-	o = append(o, 0x82, 0xad, 0x63, 0x61, 0x63, 0x68, 0x65, 0x49, 0x6e, 0x74, 0x65, 0x72, 0x76, 0x61, 0x6c)
-	o = msgp.AppendDuration(o, z.cacheInterval)
-	// string "metricsGroupOpts"
-	o = append(o, 0xb0, 0x6d, 0x65, 0x74, 0x72, 0x69, 0x63, 0x73, 0x47, 0x72, 0x6f, 0x75, 0x70, 0x4f, 0x70, 0x74, 0x73)
-	o, err = z.metricsGroupOpts.MarshalMsg(o)
-	if err != nil {
-		err = msgp.WrapError(err, "metricsGroupOpts")
-		return
-	}
-	return
-}
-
-// UnmarshalMsg implements msgp.Unmarshaler
-func (z *MetricsGroup) UnmarshalMsg(bts []byte) (o []byte, err error) {
-	var field []byte
-	_ = field
-	var zb0001 uint32
-	zb0001, bts, err = msgp.ReadMapHeaderBytes(bts)
-	if err != nil {
-		err = msgp.WrapError(err)
-		return
-	}
-	for zb0001 > 0 {
-		zb0001--
-		field, bts, err = msgp.ReadMapKeyZC(bts)
-		if err != nil {
-			err = msgp.WrapError(err)
-			return
-		}
-		switch msgp.UnsafeString(field) {
-		case "cacheInterval":
-			z.cacheInterval, bts, err = msgp.ReadDurationBytes(bts)
-			if err != nil {
-				err = msgp.WrapError(err, "cacheInterval")
-				return
-			}
-		case "metricsGroupOpts":
-			bts, err = z.metricsGroupOpts.UnmarshalMsg(bts)
-			if err != nil {
-				err = msgp.WrapError(err, "metricsGroupOpts")
-				return
-			}
-		default:
-			bts, err = msgp.Skip(bts)
-			if err != nil {
-				err = msgp.WrapError(err)
-				return
-			}
-		}
-	}
-	o = bts
-	return
-}
-
-// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
-func (z *MetricsGroup) Msgsize() (s int) {
-	s = 1 + 14 + msgp.DurationSize + 17 + z.metricsGroupOpts.Msgsize()
-	return
-}
-
 // MarshalMsg implements msgp.Marshaler
 func (z *MetricsGroupOpts) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
@@ -642,3 +577,68 @@ func (z *MetricsGroupOpts) Msgsize() (s int) {
 	s = 1 + 22 + msgp.BoolSize + 24 + msgp.BoolSize + 31 + msgp.BoolSize + 28 + msgp.BoolSize + 16 + msgp.BoolSize + 11 + msgp.BoolSize + 29 + msgp.BoolSize + 19 + msgp.BoolSize + 23 + msgp.BoolSize + 26 + msgp.BoolSize + 32 + msgp.BoolSize + 22 + msgp.BoolSize
 	return
 }
+
+// MarshalMsg implements msgp.Marshaler
+func (z *MetricsGroupV2) MarshalMsg(b []byte) (o []byte, err error) {
+	o = msgp.Require(b, z.Msgsize())
+	// map header, size 2
+	// string "cacheInterval"
+	o = append(o, 0x82, 0xad, 0x63, 0x61, 0x63, 0x68, 0x65, 0x49, 0x6e, 0x74, 0x65, 0x72, 0x76, 0x61, 0x6c)
+	o = msgp.AppendDuration(o, z.cacheInterval)
+	// string "metricsGroupOpts"
+	o = append(o, 0xb0, 0x6d, 0x65, 0x74, 0x72, 0x69, 0x63, 0x73, 0x47, 0x72, 0x6f, 0x75, 0x70, 0x4f, 0x70, 0x74, 0x73)
+	o, err = z.metricsGroupOpts.MarshalMsg(o)
+	if err != nil {
+		err = msgp.WrapError(err, "metricsGroupOpts")
+		return
+	}
+	return
+}
+
+// UnmarshalMsg implements msgp.Unmarshaler
+func (z *MetricsGroupV2) UnmarshalMsg(bts []byte) (o []byte, err error) {
+	var field []byte
+	_ = field
+	var zb0001 uint32
+	zb0001, bts, err = msgp.ReadMapHeaderBytes(bts)
+	if err != nil {
+		err = msgp.WrapError(err)
+		return
+	}
+	for zb0001 > 0 {
+		zb0001--
+		field, bts, err = msgp.ReadMapKeyZC(bts)
+		if err != nil {
+			err = msgp.WrapError(err)
+			return
+		}
+		switch msgp.UnsafeString(field) {
+		case "cacheInterval":
+			z.cacheInterval, bts, err = msgp.ReadDurationBytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "cacheInterval")
+				return
+			}
+		case "metricsGroupOpts":
+			bts, err = z.metricsGroupOpts.UnmarshalMsg(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "metricsGroupOpts")
+				return
+			}
+		default:
+			bts, err = msgp.Skip(bts)
+			if err != nil {
+				err = msgp.WrapError(err)
+				return
+			}
+		}
+	}
+	o = bts
+	return
+}
+
+// Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
+func (z *MetricsGroupV2) Msgsize() (s int) {
+	s = 1 + 14 + msgp.DurationSize + 17 + z.metricsGroupOpts.Msgsize()
+	return
+}

cmd/metrics-v2_gen_test.go

@@ -8,64 +8,6 @@ import (
 	"github.com/tinylib/msgp/msgp"
 )
 
-func TestMarshalUnmarshalMetric(t *testing.T) {
-	v := Metric{}
-	bts, err := v.MarshalMsg(nil)
-	if err != nil {
-		t.Fatal(err)
-	}
-	left, err := v.UnmarshalMsg(bts)
-	if err != nil {
-		t.Fatal(err)
-	}
-	if len(left) > 0 {
-		t.Errorf("%d bytes left over after UnmarshalMsg(): %q", len(left), left)
-	}
-
-	left, err = msgp.Skip(bts)
-	if err != nil {
-		t.Fatal(err)
-	}
-	if len(left) > 0 {
-		t.Errorf("%d bytes left over after Skip(): %q", len(left), left)
-	}
-}
-
-func BenchmarkMarshalMsgMetric(b *testing.B) {
-	v := Metric{}
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		v.MarshalMsg(nil)
-	}
-}
-
-func BenchmarkAppendMsgMetric(b *testing.B) {
-	v := Metric{}
-	bts := make([]byte, 0, v.Msgsize())
-	bts, _ = v.MarshalMsg(bts[0:0])
-	b.SetBytes(int64(len(bts)))
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		bts, _ = v.MarshalMsg(bts[0:0])
-	}
-}
-
-func BenchmarkUnmarshalMetric(b *testing.B) {
-	v := Metric{}
-	bts, _ := v.MarshalMsg(nil)
-	b.ReportAllocs()
-	b.SetBytes(int64(len(bts)))
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		_, err := v.UnmarshalMsg(bts)
-		if err != nil {
-			b.Fatal(err)
-		}
-	}
-}
-
 func TestMarshalUnmarshalMetricDescription(t *testing.T) {
 	v := MetricDescription{}
 	bts, err := v.MarshalMsg(nil)
@@ -124,8 +66,8 @@ func BenchmarkUnmarshalMetricDescription(b *testing.B) {
 	}
 }
 
-func TestMarshalUnmarshalMetricsGroup(t *testing.T) {
-	v := MetricsGroup{}
+func TestMarshalUnmarshalMetricV2(t *testing.T) {
+	v := MetricV2{}
 	bts, err := v.MarshalMsg(nil)
 	if err != nil {
 		t.Fatal(err)
@@ -147,8 +89,8 @@ func TestMarshalUnmarshalMetricsGroup(t *testing.T) {
 	}
 }
 
-func BenchmarkMarshalMsgMetricsGroup(b *testing.B) {
-	v := MetricsGroup{}
+func BenchmarkMarshalMsgMetricV2(b *testing.B) {
+	v := MetricV2{}
 	b.ReportAllocs()
 	b.ResetTimer()
 	for i := 0; i < b.N; i++ {
@@ -156,8 +98,8 @@ func BenchmarkMarshalMsgMetricsGroup(b *testing.B) {
 	}
 }
 
-func BenchmarkAppendMsgMetricsGroup(b *testing.B) {
-	v := MetricsGroup{}
+func BenchmarkAppendMsgMetricV2(b *testing.B) {
+	v := MetricV2{}
 	bts := make([]byte, 0, v.Msgsize())
 	bts, _ = v.MarshalMsg(bts[0:0])
 	b.SetBytes(int64(len(bts)))
@@ -168,8 +110,8 @@ func BenchmarkAppendMsgMetricsGroup(b *testing.B) {
 	}
 }
 
-func BenchmarkUnmarshalMetricsGroup(b *testing.B) {
-	v := MetricsGroup{}
+func BenchmarkUnmarshalMetricV2(b *testing.B) {
+	v := MetricV2{}
 	bts, _ := v.MarshalMsg(nil)
 	b.ReportAllocs()
 	b.SetBytes(int64(len(bts)))
@@ -239,3 +181,61 @@ func BenchmarkUnmarshalMetricsGroupOpts(b *testing.B) {
 		}
 	}
 }
+
+func TestMarshalUnmarshalMetricsGroupV2(t *testing.T) {
+	v := MetricsGroupV2{}
+	bts, err := v.MarshalMsg(nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	left, err := v.UnmarshalMsg(bts)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(left) > 0 {
+		t.Errorf("%d bytes left over after UnmarshalMsg(): %q", len(left), left)
+	}
+
+	left, err = msgp.Skip(bts)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(left) > 0 {
+		t.Errorf("%d bytes left over after Skip(): %q", len(left), left)
+	}
+}
+
+func BenchmarkMarshalMsgMetricsGroupV2(b *testing.B) {
+	v := MetricsGroupV2{}
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		v.MarshalMsg(nil)
+	}
+}
+
+func BenchmarkAppendMsgMetricsGroupV2(b *testing.B) {
+	v := MetricsGroupV2{}
+	bts := make([]byte, 0, v.Msgsize())
+	bts, _ = v.MarshalMsg(bts[0:0])
+	b.SetBytes(int64(len(bts)))
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		bts, _ = v.MarshalMsg(bts[0:0])
+	}
+}
+
+func BenchmarkUnmarshalMetricsGroupV2(b *testing.B) {
+	v := MetricsGroupV2{}
+	bts, _ := v.MarshalMsg(nil)
+	b.ReportAllocs()
+	b.SetBytes(int64(len(bts)))
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		_, err := v.UnmarshalMsg(bts)
+		if err != nil {
+			b.Fatal(err)
+		}
+	}
+}

cmd/metrics-v3-api.go

@@ -0,0 +1,220 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+)
+
+const (
+	apiRejectedAuthTotal      MetricName = "rejected_auth_total"
+	apiRejectedHeaderTotal    MetricName = "rejected_header_total"
+	apiRejectedTimestampTotal MetricName = "rejected_timestamp_total"
+	apiRejectedInvalidTotal   MetricName = "rejected_invalid_total"
+
+	apiRequestsWaitingTotal  MetricName = "waiting_total"
+	apiRequestsIncomingTotal MetricName = "incoming_total"
+
+	apiRequestsInFlightTotal  MetricName = "inflight_total"
+	apiRequestsTotal          MetricName = "total"
+	apiRequestsErrorsTotal    MetricName = "errors_total"
+	apiRequests5xxErrorsTotal MetricName = "5xx_errors_total"
+	apiRequests4xxErrorsTotal MetricName = "4xx_errors_total"
+	apiRequestsCanceledTotal  MetricName = "canceled_total"
+
+	apiRequestsTTFBSecondsDistribution MetricName = "ttfb_seconds_distribution"
+
+	apiTrafficSentBytes MetricName = "traffic_sent_bytes"
+	apiTrafficRecvBytes MetricName = "traffic_received_bytes"
+)
+
+var (
+	apiRejectedAuthTotalMD = NewCounterMD(apiRejectedAuthTotal,
+		"Total number of requests rejected for auth failure", "type")
+	apiRejectedHeaderTotalMD = NewCounterMD(apiRejectedHeaderTotal,
+		"Total number of requests rejected for invalid header", "type")
+	apiRejectedTimestampTotalMD = NewCounterMD(apiRejectedTimestampTotal,
+		"Total number of requests rejected for invalid timestamp", "type")
+	apiRejectedInvalidTotalMD = NewCounterMD(apiRejectedInvalidTotal,
+		"Total number of invalid requests", "type")
+
+	apiRequestsWaitingTotalMD = NewGaugeMD(apiRequestsWaitingTotal,
+		"Total number of requests in the waiting queue", "type")
+	apiRequestsIncomingTotalMD = NewGaugeMD(apiRequestsIncomingTotal,
+		"Total number of incoming requests", "type")
+
+	apiRequestsInFlightTotalMD = NewGaugeMD(apiRequestsInFlightTotal,
+		"Total number of requests currently in flight", "name", "type")
+	apiRequestsTotalMD = NewCounterMD(apiRequestsTotal,
+		"Total number of requests", "name", "type")
+	apiRequestsErrorsTotalMD = NewCounterMD(apiRequestsErrorsTotal,
+		"Total number of requests with (4xx and 5xx) errors", "name", "type")
+	apiRequests5xxErrorsTotalMD = NewCounterMD(apiRequests5xxErrorsTotal,
+		"Total number of requests with 5xx errors", "name", "type")
+	apiRequests4xxErrorsTotalMD = NewCounterMD(apiRequests4xxErrorsTotal,
+		"Total number of requests with 4xx errors", "name", "type")
+	apiRequestsCanceledTotalMD = NewCounterMD(apiRequestsCanceledTotal,
+		"Total number of requests canceled by the client", "name", "type")
+
+	apiRequestsTTFBSecondsDistributionMD = NewCounterMD(apiRequestsTTFBSecondsDistribution,
+		"Distribution of time to first byte across API calls", "name", "type", "le")
+
+	apiTrafficSentBytesMD = NewCounterMD(apiTrafficSentBytes,
+		"Total number of bytes sent", "type")
+	apiTrafficRecvBytesMD = NewCounterMD(apiTrafficRecvBytes,
+		"Total number of bytes received", "type")
+)
+
+// loadAPIRequestsHTTPMetrics - reads S3 HTTP metrics.
+//
+// This is a `MetricsLoaderFn`.
+//
+// This includes node level S3 HTTP metrics.
+//
+// This function currently ignores `opts`.
+func loadAPIRequestsHTTPMetrics(ctx context.Context, m MetricValues, _ *metricsCache) error {
+	// Collect node level S3 HTTP metrics.
+	httpStats := globalHTTPStats.toServerHTTPStats(false)
+
+	// Currently we only collect S3 API related stats, so we set the "type"
+	// label to "s3".
+
+	m.Set(apiRejectedAuthTotal, float64(httpStats.TotalS3RejectedAuth), "type", "s3")
+	m.Set(apiRejectedTimestampTotal, float64(httpStats.TotalS3RejectedTime), "type", "s3")
+	m.Set(apiRejectedHeaderTotal, float64(httpStats.TotalS3RejectedHeader), "type", "s3")
+	m.Set(apiRejectedInvalidTotal, float64(httpStats.TotalS3RejectedInvalid), "type", "s3")
+	m.Set(apiRequestsWaitingTotal, float64(httpStats.S3RequestsInQueue), "type", "s3")
+	m.Set(apiRequestsIncomingTotal, float64(httpStats.S3RequestsIncoming), "type", "s3")
+
+	for name, value := range httpStats.CurrentS3Requests.APIStats {
+		m.Set(apiRequestsInFlightTotal, float64(value), "name", name, "type", "s3")
+	}
+	for name, value := range httpStats.TotalS3Requests.APIStats {
+		m.Set(apiRequestsTotal, float64(value), "name", name, "type", "s3")
+	}
+	for name, value := range httpStats.TotalS3Errors.APIStats {
+		m.Set(apiRequestsErrorsTotal, float64(value), "name", name, "type", "s3")
+	}
+	for name, value := range httpStats.TotalS35xxErrors.APIStats {
+		m.Set(apiRequests5xxErrorsTotal, float64(value), "name", name, "type", "s3")
+	}
+	for name, value := range httpStats.TotalS34xxErrors.APIStats {
+		m.Set(apiRequests4xxErrorsTotal, float64(value), "name", name, "type", "s3")
+	}
+	for name, value := range httpStats.TotalS3Canceled.APIStats {
+		m.Set(apiRequestsCanceledTotal, float64(value), "name", name, "type", "s3")
+	}
+	return nil
+}
+
+// loadAPIRequestsTTFBMetrics - loads S3 TTFB metrics.
+//
+// This is a `MetricsLoaderFn`.
+func loadAPIRequestsTTFBMetrics(ctx context.Context, m MetricValues, _ *metricsCache) error {
+	renameLabels := map[string]string{"api": "name"}
+	m.SetHistogram(apiRequestsTTFBSecondsDistribution, httpRequestsDuration, renameLabels, nil,
+		"type", "s3")
+	return nil
+}
+
+// loadAPIRequestsNetworkMetrics - loads S3 network metrics.
+//
+// This is a `MetricsLoaderFn`.
+func loadAPIRequestsNetworkMetrics(ctx context.Context, m MetricValues, _ *metricsCache) error {
+	connStats := globalConnStats.toServerConnStats()
+	m.Set(apiTrafficSentBytes, float64(connStats.s3OutputBytes), "type", "s3")
+	m.Set(apiTrafficRecvBytes, float64(connStats.s3InputBytes), "type", "s3")
+	return nil
+}
+
+// Metric Descriptions for bucket level S3 metrics.
+var (
+	apiBucketTrafficSentBytesMD = NewCounterMD(apiTrafficSentBytes,
+		"Total number of bytes received for a bucket", "bucket", "type")
+	apiBucketTrafficRecvBytesMD = NewCounterMD(apiTrafficRecvBytes,
+		"Total number of bytes sent for a bucket", "bucket", "type")
+
+	apiBucketRequestsInFlightMD = NewGaugeMD(apiRequestsInFlightTotal,
+		"Total number of requests currently in flight for a bucket", "bucket", "name", "type")
+	apiBucketRequestsTotalMD = NewCounterMD(apiRequestsTotal,
+		"Total number of requests for a bucket", "bucket", "name", "type")
+	apiBucketRequestsCanceledMD = NewCounterMD(apiRequestsCanceledTotal,
+		"Total number of requests canceled by the client for a bucket", "bucket", "name", "type")
+	apiBucketRequests4xxErrorsMD = NewCounterMD(apiRequests4xxErrorsTotal,
+		"Total number of requests with 4xx errors for a bucket", "bucket", "name", "type")
+	apiBucketRequests5xxErrorsMD = NewCounterMD(apiRequests5xxErrorsTotal,
+		"Total number of requests with 5xx errors for a bucket", "bucket", "name", "type")
+
+	apiBucketRequestsTTFBSecondsDistributionMD = NewCounterMD(apiRequestsTTFBSecondsDistribution,
+		"Distribution of time to first byte across API calls for a bucket",
+		"bucket", "name", "le", "type")
+)
+
+// loadAPIBucketHTTPMetrics - loads bucket level S3 HTTP metrics.
+//
+// This is a `MetricsLoaderFn`.
+//
+// This includes bucket level S3 HTTP metrics and S3 network in/out metrics.
+func loadAPIBucketHTTPMetrics(ctx context.Context, m MetricValues, _ *metricsCache, buckets []string) error {
+	if len(buckets) == 0 {
+		return nil
+	}
+	for bucket, inOut := range globalBucketConnStats.getBucketS3InOutBytes(buckets) {
+		recvBytes := inOut.In
+		if recvBytes > 0 {
+			m.Set(apiTrafficSentBytes, float64(recvBytes), "bucket", bucket, "type", "s3")
+		}
+		sentBytes := inOut.Out
+		if sentBytes > 0 {
+			m.Set(apiTrafficRecvBytes, float64(sentBytes), "bucket", bucket, "type", "s3")
+		}
+
+		httpStats := globalBucketHTTPStats.load(bucket)
+		for k, v := range httpStats.currentS3Requests.Load(false) {
+			m.Set(apiRequestsInFlightTotal, float64(v), "bucket", bucket, "name", k, "type", "s3")
+		}
+
+		for k, v := range httpStats.totalS3Requests.Load(false) {
+			m.Set(apiRequestsTotal, float64(v), "bucket", bucket, "name", k, "type", "s3")
+		}
+
+		for k, v := range httpStats.totalS3Canceled.Load(false) {
+			m.Set(apiRequestsCanceledTotal, float64(v), "bucket", bucket, "name", k, "type", "s3")
+		}
+
+		for k, v := range httpStats.totalS34xxErrors.Load(false) {
+			m.Set(apiRequests4xxErrorsTotal, float64(v), "bucket", bucket, "name", k, "type", "s3")
+		}
+
+		for k, v := range httpStats.totalS35xxErrors.Load(false) {
+			m.Set(apiRequests5xxErrorsTotal, float64(v), "bucket", bucket, "name", k, "type", "s3")
+		}
+	}
+
+	return nil
+}
+
+// loadAPIBucketTTFBMetrics - loads bucket S3 TTFB metrics.
+//
+// This is a `MetricsLoaderFn`.
+func loadAPIBucketTTFBMetrics(ctx context.Context, m MetricValues, _ *metricsCache, buckets []string) error {
+	renameLabels := map[string]string{"api": "name"}
+	m.SetHistogram(apiRequestsTTFBSecondsDistribution, bucketHTTPRequestsDuration, renameLabels,
+		buckets, "type", "s3")
+	return nil
+}

cmd/metrics-v3-cache.go

@@ -0,0 +1,145 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"time"
+
+	"github.com/minio/madmin-go/v3"
+	"github.com/minio/minio/internal/cachevalue"
+)
+
+// metricsCache - cache for metrics.
+//
+// When serving metrics, this cache is passed to the MetricsLoaderFn.
+//
+// This cache is used for metrics that would result in network/storage calls.
+type metricsCache struct {
+	dataUsageInfo       *cachevalue.Cache[DataUsageInfo]
+	esetHealthResult    *cachevalue.Cache[HealthResult]
+	driveMetrics        *cachevalue.Cache[storageMetrics]
+	clusterDriveMetrics *cachevalue.Cache[storageMetrics]
+	nodesUpDown         *cachevalue.Cache[nodesOnline]
+}
+
+func newMetricsCache() *metricsCache {
+	return &metricsCache{
+		dataUsageInfo:       newDataUsageInfoCache(),
+		esetHealthResult:    newESetHealthResultCache(),
+		driveMetrics:        newDriveMetricsCache(),
+		clusterDriveMetrics: newClusterStorageInfoCache(),
+		nodesUpDown:         newNodesUpDownCache(),
+	}
+}
+
+type nodesOnline struct {
+	Online, Offline int
+}
+
+func newNodesUpDownCache() *cachevalue.Cache[nodesOnline] {
+	loadNodesUpDown := func() (v nodesOnline, err error) {
+		v.Online, v.Offline = globalNotificationSys.GetPeerOnlineCount()
+		return
+	}
+	return cachevalue.NewFromFunc(1*time.Minute,
+		cachevalue.Opts{ReturnLastGood: true},
+		loadNodesUpDown)
+}
+
+type storageMetrics struct {
+	storageInfo                              madmin.StorageInfo
+	onlineDrives, offlineDrives, totalDrives int
+}
+
+func newDataUsageInfoCache() *cachevalue.Cache[DataUsageInfo] {
+	loadDataUsage := func() (u DataUsageInfo, err error) {
+		objLayer := newObjectLayerFn()
+		if objLayer == nil {
+			return
+		}
+
+		// Collect cluster level object metrics.
+		u, err = loadDataUsageFromBackend(GlobalContext, objLayer)
+		return
+	}
+	return cachevalue.NewFromFunc(1*time.Minute,
+		cachevalue.Opts{ReturnLastGood: true},
+		loadDataUsage)
+}
+
+func newESetHealthResultCache() *cachevalue.Cache[HealthResult] {
+	loadHealth := func() (r HealthResult, err error) {
+		objLayer := newObjectLayerFn()
+		if objLayer == nil {
+			return
+		}
+
+		r = objLayer.Health(GlobalContext, HealthOptions{})
+		return
+	}
+	return cachevalue.NewFromFunc(1*time.Minute,
+		cachevalue.Opts{ReturnLastGood: true},
+		loadHealth,
+	)
+}
+
+func newDriveMetricsCache() *cachevalue.Cache[storageMetrics] {
+	loadDriveMetrics := func() (v storageMetrics, err error) {
+		objLayer := newObjectLayerFn()
+		if objLayer == nil {
+			return
+		}
+
+		storageInfo := objLayer.LocalStorageInfo(GlobalContext, true)
+		onlineDrives, offlineDrives := getOnlineOfflineDisksStats(storageInfo.Disks)
+		totalDrives := onlineDrives.Merge(offlineDrives)
+		v = storageMetrics{
+			storageInfo:   storageInfo,
+			onlineDrives:  onlineDrives.Sum(),
+			offlineDrives: offlineDrives.Sum(),
+			totalDrives:   totalDrives.Sum(),
+		}
+		return
+	}
+	return cachevalue.NewFromFunc(1*time.Minute,
+		cachevalue.Opts{ReturnLastGood: true},
+		loadDriveMetrics)
+}
+
+func newClusterStorageInfoCache() *cachevalue.Cache[storageMetrics] {
+	loadStorageInfo := func() (v storageMetrics, err error) {
+		objLayer := newObjectLayerFn()
+		if objLayer == nil {
+			return storageMetrics{}, nil
+		}
+		storageInfo := objLayer.StorageInfo(GlobalContext, true)
+		onlineDrives, offlineDrives := getOnlineOfflineDisksStats(storageInfo.Disks)
+		totalDrives := onlineDrives.Merge(offlineDrives)
+		v = storageMetrics{
+			storageInfo:   storageInfo,
+			onlineDrives:  onlineDrives.Sum(),
+			offlineDrives: offlineDrives.Sum(),
+			totalDrives:   totalDrives.Sum(),
+		}
+		return
+	}
+	return cachevalue.NewFromFunc(1*time.Minute,
+		cachevalue.Opts{ReturnLastGood: true},
+		loadStorageInfo,
+	)
+}

cmd/metrics-v3-cluster-erasure-set.go

@@ -0,0 +1,89 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+	"strconv"
+)
+
+const (
+	erasureSetOverallWriteQuorum = "overall_write_quorum"
+	erasureSetOverallHealth      = "overall_health"
+	erasureSetReadQuorum         = "read_quorum"
+	erasureSetWriteQuorum        = "write_quorum"
+	erasureSetOnlineDrivesCount  = "online_drives_count"
+	erasureSetHealingDrivesCount = "healing_drives_count"
+	erasureSetHealth             = "health"
+)
+
+const (
+	poolIDL = "pool_id"
+	setIDL  = "set_id"
+)
+
+var (
+	erasureSetOverallWriteQuorumMD = NewGaugeMD(erasureSetOverallWriteQuorum,
+		"Overall write quorum across pools and sets")
+	erasureSetOverallHealthMD = NewGaugeMD(erasureSetOverallHealth,
+		"Overall health across pools and sets (1=healthy, 0=unhealthy)")
+	erasureSetReadQuorumMD = NewGaugeMD(erasureSetReadQuorum,
+		"Read quorum for the erasure set in a pool", poolIDL, setIDL)
+	erasureSetWriteQuorumMD = NewGaugeMD(erasureSetWriteQuorum,
+		"Write quorum for the erasure set in a pool", poolIDL, setIDL)
+	erasureSetOnlineDrivesCountMD = NewGaugeMD(erasureSetOnlineDrivesCount,
+		"Count of online drives in the erasure set in a pool", poolIDL, setIDL)
+	erasureSetHealingDrivesCountMD = NewGaugeMD(erasureSetHealingDrivesCount,
+		"Count of healing drives in the erasure set in a pool", poolIDL, setIDL)
+	erasureSetHealthMD = NewGaugeMD(erasureSetHealth,
+		"Health of the erasure set in a pool (1=healthy, 0=unhealthy)",
+		poolIDL, setIDL)
+)
+
+func b2f(v bool) float64 {
+	if v {
+		return 1
+	}
+	return 0
+}
+
+// loadClusterErasureSetMetrics - `MetricsLoaderFn` for cluster storage erasure
+// set metrics.
+func loadClusterErasureSetMetrics(ctx context.Context, m MetricValues, c *metricsCache) error {
+	result, _ := c.esetHealthResult.Get()
+
+	m.Set(erasureSetOverallWriteQuorum, float64(result.WriteQuorum))
+	m.Set(erasureSetOverallHealth, b2f(result.Healthy))
+
+	for _, h := range result.ESHealth {
+		poolLV := strconv.Itoa(h.PoolID)
+		setLV := strconv.Itoa(h.SetID)
+		m.Set(erasureSetReadQuorum, float64(h.ReadQuorum),
+			poolIDL, poolLV, setIDL, setLV)
+		m.Set(erasureSetWriteQuorum, float64(h.WriteQuorum),
+			poolIDL, poolLV, setIDL, setLV)
+		m.Set(erasureSetOnlineDrivesCount, float64(h.HealthyDrives),
+			poolIDL, poolLV, setIDL, setLV)
+		m.Set(erasureSetHealingDrivesCount, float64(h.HealingDrives),
+			poolIDL, poolLV, setIDL, setLV)
+		m.Set(erasureSetHealth, b2f(h.Healthy),
+			poolIDL, poolLV, setIDL, setLV)
+	}
+
+	return nil
+}

cmd/metrics-v3-cluster-health.go

@@ -0,0 +1,109 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import "context"
+
+const (
+	healthDrivesOfflineCount = "drives_offline_count"
+	healthDrivesOnlineCount  = "drives_online_count"
+	healthDrivesCount        = "drives_count"
+)
+
+var (
+	healthDrivesOfflineCountMD = NewGaugeMD(healthDrivesOfflineCount,
+		"Count of offline drives in the cluster")
+	healthDrivesOnlineCountMD = NewGaugeMD(healthDrivesOnlineCount,
+		"Count of online drives in the cluster")
+	healthDrivesCountMD = NewGaugeMD(healthDrivesCount,
+		"Count of all drives in the cluster")
+)
+
+// loadClusterHealthDriveMetrics - `MetricsLoaderFn` for cluster storage drive metrics
+// such as online, offline and total drives.
+func loadClusterHealthDriveMetrics(ctx context.Context, m MetricValues,
+	c *metricsCache,
+) error {
+	clusterDriveMetrics, _ := c.clusterDriveMetrics.Get()
+
+	m.Set(healthDrivesOfflineCount, float64(clusterDriveMetrics.offlineDrives))
+	m.Set(healthDrivesOnlineCount, float64(clusterDriveMetrics.onlineDrives))
+	m.Set(healthDrivesCount, float64(clusterDriveMetrics.totalDrives))
+
+	return nil
+}
+
+const (
+	healthNodesOfflineCount = "nodes_offline_count"
+	healthNodesOnlineCount  = "nodes_online_count"
+)
+
+var (
+	healthNodesOfflineCountMD = NewGaugeMD(healthNodesOfflineCount,
+		"Count of offline nodes in the cluster")
+	healthNodesOnlineCountMD = NewGaugeMD(healthNodesOnlineCount,
+		"Count of online nodes in the cluster")
+)
+
+// loadClusterHealthNodeMetrics - `MetricsLoaderFn` for cluster health node
+// metrics.
+func loadClusterHealthNodeMetrics(ctx context.Context, m MetricValues,
+	c *metricsCache,
+) error {
+	nodesUpDown, _ := c.nodesUpDown.Get()
+
+	m.Set(healthNodesOfflineCount, float64(nodesUpDown.Offline))
+	m.Set(healthNodesOnlineCount, float64(nodesUpDown.Online))
+
+	return nil
+}
+
+const (
+	healthCapacityRawTotalBytes    = "capacity_raw_total_bytes"
+	healthCapacityRawFreeBytes     = "capacity_raw_free_bytes"
+	healthCapacityUsableTotalBytes = "capacity_usable_total_bytes"
+	healthCapacityUsableFreeBytes  = "capacity_usable_free_bytes"
+)
+
+var (
+	healthCapacityRawTotalBytesMD = NewGaugeMD(healthCapacityRawTotalBytes,
+		"Total cluster raw storage capacity in bytes")
+	healthCapacityRawFreeBytesMD = NewGaugeMD(healthCapacityRawFreeBytes,
+		"Total cluster raw storage free in bytes")
+	healthCapacityUsableTotalBytesMD = NewGaugeMD(healthCapacityUsableTotalBytes,
+		"Total cluster usable storage capacity in bytes")
+	healthCapacityUsableFreeBytesMD = NewGaugeMD(healthCapacityUsableFreeBytes,
+		"Total cluster usable storage free in bytes")
+)
+
+// loadClusterHealthCapacityMetrics - `MetricsLoaderFn` for cluster storage
+// capacity metrics.
+func loadClusterHealthCapacityMetrics(ctx context.Context, m MetricValues,
+	c *metricsCache,
+) error {
+	clusterDriveMetrics, _ := c.clusterDriveMetrics.Get()
+
+	storageInfo := clusterDriveMetrics.storageInfo
+
+	m.Set(healthCapacityRawTotalBytes, float64(GetTotalCapacity(storageInfo.Disks)))
+	m.Set(healthCapacityRawFreeBytes, float64(GetTotalCapacityFree(storageInfo.Disks)))
+	m.Set(healthCapacityUsableTotalBytes, float64(GetTotalUsableCapacity(storageInfo.Disks, storageInfo)))
+	m.Set(healthCapacityUsableFreeBytes, float64(GetTotalUsableCapacityFree(storageInfo.Disks, storageInfo)))
+
+	return nil
+}

cmd/metrics-v3-cluster-usage.go

@@ -0,0 +1,189 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+	"time"
+
+	"github.com/minio/minio/internal/logger"
+)
+
+const (
+	usageSinceLastUpdateSeconds   = "since_last_update_seconds"
+	usageTotalBytes               = "total_bytes"
+	usageObjectsCount             = "count"
+	usageVersionsCount            = "versions_count"
+	usageDeleteMarkersCount       = "delete_markers_count"
+	usageBucketsCount             = "buckets_count"
+	usageSizeDistribution         = "size_distribution"
+	usageVersionCountDistribution = "version_count_distribution"
+)
+
+var (
+	usageSinceLastUpdateSecondsMD = NewGaugeMD(usageSinceLastUpdateSeconds,
+		"Time since last update of usage metrics in seconds")
+	usageTotalBytesMD = NewGaugeMD(usageTotalBytes,
+		"Total cluster usage in bytes")
+	usageObjectsCountMD = NewGaugeMD(usageObjectsCount,
+		"Total cluster objects count")
+	usageVersionsCountMD = NewGaugeMD(usageVersionsCount,
+		"Total cluster object versions (including delete markers) count")
+	usageDeleteMarkersCountMD = NewGaugeMD(usageDeleteMarkersCount,
+		"Total cluster delete markers count")
+	usageBucketsCountMD = NewGaugeMD(usageBucketsCount,
+		"Total cluster buckets count")
+	usageObjectsDistributionMD = NewGaugeMD(usageSizeDistribution,
+		"Cluster object size distribution", "range")
+	usageVersionsDistributionMD = NewGaugeMD(usageVersionCountDistribution,
+		"Cluster object version count distribution", "range")
+)
+
+// loadClusterUsageObjectMetrics - reads cluster usage metrics.
+//
+// This is a `MetricsLoaderFn`.
+func loadClusterUsageObjectMetrics(ctx context.Context, m MetricValues, c *metricsCache) error {
+	dataUsageInfo, err := c.dataUsageInfo.Get()
+	if err != nil {
+		logger.LogIf(ctx, err)
+		return nil
+	}
+
+	// data usage has not captured any data yet.
+	if dataUsageInfo.LastUpdate.IsZero() {
+		return nil
+	}
+
+	var (
+		clusterSize               uint64
+		clusterBuckets            uint64
+		clusterObjectsCount       uint64
+		clusterVersionsCount      uint64
+		clusterDeleteMarkersCount uint64
+	)
+
+	clusterObjectSizesHistogram := map[string]uint64{}
+	clusterVersionsHistogram := map[string]uint64{}
+	for _, usage := range dataUsageInfo.BucketsUsage {
+		clusterBuckets++
+		clusterSize += usage.Size
+		clusterObjectsCount += usage.ObjectsCount
+		clusterVersionsCount += usage.VersionsCount
+		clusterDeleteMarkersCount += usage.DeleteMarkersCount
+		for k, v := range usage.ObjectSizesHistogram {
+			clusterObjectSizesHistogram[k] += v
+		}
+		for k, v := range usage.ObjectVersionsHistogram {
+			clusterVersionsHistogram[k] += v
+		}
+	}
+
+	m.Set(usageSinceLastUpdateSeconds, time.Since(dataUsageInfo.LastUpdate).Seconds())
+	m.Set(usageTotalBytes, float64(clusterSize))
+	m.Set(usageObjectsCount, float64(clusterObjectsCount))
+	m.Set(usageVersionsCount, float64(clusterVersionsCount))
+	m.Set(usageDeleteMarkersCount, float64(clusterDeleteMarkersCount))
+	m.Set(usageBucketsCount, float64(clusterBuckets))
+	for k, v := range clusterObjectSizesHistogram {
+		m.Set(usageSizeDistribution, float64(v), "range", k)
+	}
+	for k, v := range clusterVersionsHistogram {
+		m.Set(usageVersionCountDistribution, float64(v), "range", k)
+	}
+
+	return nil
+}
+
+const (
+	usageBucketQuotaTotalBytes = "quota_total_bytes"
+
+	usageBucketTotalBytes                     = "total_bytes"
+	usageBucketObjectsCount                   = "objects_count"
+	usageBucketVersionsCount                  = "versions_count"
+	usageBucketDeleteMarkersCount             = "delete_markers_count"
+	usageBucketObjectSizeDistribution         = "object_size_distribution"
+	usageBucketObjectVersionCountDistribution = "object_version_count_distribution"
+)
+
+var (
+	usageBucketTotalBytesMD = NewGaugeMD(usageBucketTotalBytes,
+		"Total bucket size in bytes", "bucket")
+	usageBucketObjectsTotalMD = NewGaugeMD(usageBucketObjectsCount,
+		"Total objects count in bucket", "bucket")
+	usageBucketVersionsCountMD = NewGaugeMD(usageBucketVersionsCount,
+		"Total object versions (including delete markers) count in bucket", "bucket")
+	usageBucketDeleteMarkersCountMD = NewGaugeMD(usageBucketDeleteMarkersCount,
+		"Total delete markers count in bucket", "bucket")
+
+	usageBucketQuotaTotalBytesMD = NewGaugeMD(usageBucketQuotaTotalBytes,
+		"Total bucket quota in bytes", "bucket")
+
+	usageBucketObjectSizeDistributionMD = NewGaugeMD(usageBucketObjectSizeDistribution,
+		"Bucket object size distribution", "range", "bucket")
+	usageBucketObjectVersionCountDistributionMD = NewGaugeMD(
+		usageBucketObjectVersionCountDistribution,
+		"Bucket object version count distribution", "range", "bucket")
+)
+
+// loadClusterUsageBucketMetrics - `MetricsLoaderFn` to load bucket usage metrics.
+func loadClusterUsageBucketMetrics(ctx context.Context, m MetricValues, c *metricsCache, buckets []string) error {
+	dataUsageInfo, err := c.dataUsageInfo.Get()
+	if err != nil {
+		logger.LogIf(ctx, err)
+		return nil
+	}
+
+	// data usage has not been captured yet.
+	if dataUsageInfo.LastUpdate.IsZero() {
+		return nil
+	}
+
+	m.Set(usageSinceLastUpdateSeconds, float64(time.Since(dataUsageInfo.LastUpdate)))
+
+	for _, bucket := range buckets {
+		usage, ok := dataUsageInfo.BucketsUsage[bucket]
+		if !ok {
+			continue
+		}
+		quota, err := globalBucketQuotaSys.Get(ctx, bucket)
+		if err != nil {
+			// Log and continue if we are unable to retrieve metrics for this
+			// bucket.
+			logger.LogIf(ctx, err)
+			continue
+		}
+
+		m.Set(usageBucketTotalBytes, float64(usage.Size), "bucket", bucket)
+		m.Set(usageBucketObjectsCount, float64(usage.ObjectsCount), "bucket", bucket)
+		m.Set(usageBucketVersionsCount, float64(usage.VersionsCount), "bucket", bucket)
+		m.Set(usageBucketDeleteMarkersCount, float64(usage.DeleteMarkersCount), "bucket", bucket)
+
+		if quota != nil && quota.Quota > 0 {
+			m.Set(usageBucketQuotaTotalBytes, float64(quota.Quota), "bucket", bucket)
+		}
+
+		for k, v := range usage.ObjectSizesHistogram {
+			m.Set(usageBucketObjectSizeDistribution, float64(v), "range", k, "bucket", bucket)
+		}
+		for k, v := range usage.ObjectVersionsHistogram {
+			m.Set(usageBucketObjectVersionCountDistribution, float64(v), "range", k, "bucket", bucket)
+		}
+
+	}
+	return nil
+}

cmd/metrics-v3-handler.go

@@ -0,0 +1,254 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"slices"
+	"strings"
+
+	"github.com/minio/minio/internal/logger"
+	"github.com/minio/minio/internal/mcontext"
+	"github.com/minio/mux"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
+)
+
+type promLogger struct{}
+
+func (p promLogger) Println(v ...interface{}) {
+	s := make([]string, 0, len(v))
+	for _, val := range v {
+		s = append(s, fmt.Sprintf("%v", val))
+	}
+	err := fmt.Errorf("metrics handler error: %v", strings.Join(s, " "))
+	logger.LogIf(GlobalContext, err)
+}
+
+type metricsV3Server struct {
+	registry *prometheus.Registry
+	opts     promhttp.HandlerOpts
+	authFn   func(http.Handler) http.Handler
+
+	metricsData *metricsV3Collection
+}
+
+func newMetricsV3Server(authType prometheusAuthType) *metricsV3Server {
+	registry := prometheus.NewRegistry()
+	authFn := AuthMiddleware
+	if authType == prometheusPublic {
+		authFn = NoAuthMiddleware
+	}
+
+	metricGroups := newMetricGroups(registry)
+
+	return &metricsV3Server{
+		registry: registry,
+		opts: promhttp.HandlerOpts{
+			ErrorLog:            promLogger{},
+			ErrorHandling:       promhttp.HTTPErrorOnError,
+			Registry:            registry,
+			MaxRequestsInFlight: 2,
+		},
+		authFn: authFn,
+
+		metricsData: metricGroups,
+	}
+}
+
+// metricDisplay - contains info on a metric for display purposes.
+type metricDisplay struct {
+	Name   string   `json:"name"`
+	Help   string   `json:"help"`
+	Type   string   `json:"type"`
+	Labels []string `json:"labels"`
+}
+
+func (md metricDisplay) String() string {
+	return fmt.Sprintf("Name: %s\nType: %s\nHelp: %s\nLabels: {%s}\n", md.Name, md.Type, md.Help, strings.Join(md.Labels, ","))
+}
+
+func (md metricDisplay) TableRow() string {
+	labels := strings.Join(md.Labels, ",")
+	if labels == "" {
+		labels = ""
+	} else {
+		labels = "`" + labels + "`"
+	}
+	return fmt.Sprintf("| `%s` | `%s` | %s | %s |\n", md.Name, md.Type, md.Help, labels)
+}
+
+// listMetrics - returns a handler that lists all the metrics that could be
+// returned for the requested path.
+//
+// FIXME: It currently only lists `minio_` prefixed metrics.
+func (h *metricsV3Server) listMetrics(path string) http.Handler {
+	// First collect all matching MetricsGroup's
+	matchingMG := make(map[collectorPath]*MetricsGroup)
+	for _, collPath := range h.metricsData.collectorPaths {
+		if collPath.isDescendantOf(path) {
+			if v, ok := h.metricsData.mgMap[collPath]; ok {
+				matchingMG[collPath] = v
+			} else {
+				matchingMG[collPath] = h.metricsData.bucketMGMap[collPath]
+			}
+		}
+	}
+
+	if len(matchingMG) == 0 {
+		return nil
+	}
+
+	var metrics []metricDisplay
+	for _, collectorPath := range h.metricsData.collectorPaths {
+		if mg, ok := matchingMG[collectorPath]; ok {
+			var commonLabels []string
+			for k := range mg.ExtraLabels {
+				commonLabels = append(commonLabels, k)
+			}
+			for _, d := range mg.Descriptors {
+				labels := slices.Clone(d.VariableLabels)
+				labels = append(labels, commonLabels...)
+				metric := metricDisplay{
+					Name:   mg.MetricFQN(d.Name),
+					Help:   d.Help,
+					Type:   d.Type.String(),
+					Labels: labels,
+				}
+				metrics = append(metrics, metric)
+			}
+		}
+	}
+
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		contentType := r.Header.Get("Content-Type")
+		if contentType == "application/json" {
+			w.Header().Set("Content-Type", "application/json")
+			jsonEncoder := json.NewEncoder(w)
+			jsonEncoder.Encode(metrics)
+			return
+		}
+
+		// If not JSON, return plain text. We format it as a markdown table for
+		// readability.
+		w.Header().Set("Content-Type", "text/plain")
+		var b strings.Builder
+		b.WriteString("| Name | Type | Help | Labels |\n")
+		b.WriteString("| ---- | ---- | ---- | ------ |\n")
+		for _, metric := range metrics {
+			b.WriteString(metric.TableRow())
+		}
+		w.Write([]byte(b.String()))
+	})
+}
+
+func (h *metricsV3Server) handle(path string, isListingRequest bool, buckets []string) http.Handler {
+	var notFoundHandler http.Handler = http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		http.Error(w, "Metrics Resource Not found", http.StatusNotFound)
+	})
+
+	// Require that metrics path has at least component.
+	if path == "/" {
+		return notFoundHandler
+	}
+
+	if isListingRequest {
+		handler := h.listMetrics(path)
+		if handler == nil {
+			return notFoundHandler
+		}
+		return handler
+	}
+
+	// In each of the following cases, we check if the collect path is a
+	// descendant of `path`, and if so, we add the corresponding gatherer to
+	// the list of gatherers. This way, /api/a will return all metrics returned
+	// by /api/a/b and /api/a/c (and any other matching descendant collector
+	// paths).
+
+	var gatherers []prometheus.Gatherer
+	for _, collectorPath := range h.metricsData.collectorPaths {
+		if collectorPath.isDescendantOf(path) {
+			gatherer := h.metricsData.mgGatherers[collectorPath]
+
+			// For Bucket metrics we need to set the buckets argument inside the
+			// metric group, so that it will affect collection. If no buckets
+			// are provided, we will not return bucket metrics.
+			if bmg, ok := h.metricsData.bucketMGMap[collectorPath]; ok {
+				if len(buckets) == 0 {
+					continue
+				}
+				unLocker := bmg.LockAndSetBuckets(buckets)
+				defer unLocker()
+			}
+			gatherers = append(gatherers, gatherer)
+		}
+	}
+
+	if len(gatherers) == 0 {
+		return notFoundHandler
+	}
+
+	return promhttp.HandlerFor(prometheus.Gatherers(gatherers), h.opts)
+}
+
+// ServeHTTP - implements http.Handler interface.
+//
+// When the `list` query parameter is provided (its value is ignored), the
+// server lists all metrics that could be returned for the requested path.
+//
+// The (repeatable) `buckets` query parameter is a list of bucket names (or it
+// could be a comma separated value) to return metrics with a bucket label.
+// Bucket metrics will be returned only for the provided buckets. If no buckets
+// parameter is provided, no bucket metrics are returned.
+func (h *metricsV3Server) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+	pathComponents := mux.Vars(r)["pathComps"]
+	isListingRequest := r.Form.Has("list")
+
+	// Parse optional buckets query parameter.
+	bucketsParam := r.Form["buckets"]
+	buckets := make([]string, 0, len(bucketsParam))
+	for _, bp := range bucketsParam {
+		bp = strings.TrimSpace(bp)
+		if bp == "" {
+			continue
+		}
+		splits := strings.Split(bp, ",")
+		for _, split := range splits {
+			buckets = append(buckets, strings.TrimSpace(split))
+		}
+	}
+
+	innerHandler := h.handle(pathComponents, isListingRequest, buckets)
+
+	// Add tracing to the prom. handler
+	tracedHandler := http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		tc, ok := r.Context().Value(mcontext.ContextTraceKey).(*mcontext.TraceCtxt)
+		if ok {
+			tc.FuncName = "handler.MetricsV3"
+			tc.ResponseRecorder.LogErrBody = true
+		}
+
+		innerHandler.ServeHTTP(w, r)
+	})
+
+	// Add authentication
+	h.authFn(tracedHandler).ServeHTTP(w, r)
+}

cmd/metrics-v3-system-drive.go

@@ -0,0 +1,126 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+	"strconv"
+
+	"github.com/minio/minio/internal/logger"
+)
+
+// label constants
+const (
+	driveL      = "drive"
+	poolIndexL  = "pool_index"
+	setIndexL   = "set_index"
+	driveIndexL = "drive_index"
+
+	apiL = "api"
+)
+
+var allDriveLabels = []string{driveL, poolIndexL, setIndexL, driveIndexL}
+
+const (
+	driveUsedBytes               = "used_bytes"
+	driveFreeBytes               = "free_bytes"
+	driveTotalBytes              = "total_bytes"
+	driveFreeInodes              = "free_inodes"
+	driveTimeoutErrorsTotal      = "timeout_errors_total"
+	driveAvailabilityErrorsTotal = "availability_errors_total"
+	driveWaitingIO               = "waiting_io"
+	driveAPILatencyMicros        = "api_latency_micros"
+
+	driveOfflineCount = "offline_count"
+	driveOnlineCount  = "online_count"
+	driveCount        = "count"
+)
+
+var (
+	driveUsedBytesMD = NewGaugeMD(driveUsedBytes,
+		"Total storage used on a drive in bytes", allDriveLabels...)
+	driveFreeBytesMD = NewGaugeMD(driveFreeBytes,
+		"Total storage free on a drive in bytes", allDriveLabels...)
+	driveTotalBytesMD = NewGaugeMD(driveTotalBytes,
+		"Total storage available on a drive in bytes", allDriveLabels...)
+	driveFreeInodesMD = NewGaugeMD(driveFreeInodes,
+		"Total free inodes on a drive", allDriveLabels...)
+	driveTimeoutErrorsMD = NewCounterMD(driveTimeoutErrorsTotal,
+		"Total timeout errors on a drive", allDriveLabels...)
+	driveAvailabilityErrorsMD = NewCounterMD(driveAvailabilityErrorsTotal,
+		"Total availability errors (I/O errors, permission denied and timeouts) on a drive",
+		allDriveLabels...)
+	driveWaitingIOMD = NewGaugeMD(driveWaitingIO,
+		"Total waiting I/O operations on a drive", allDriveLabels...)
+	driveAPILatencyMD = NewGaugeMD(driveAPILatencyMicros,
+		"Average last minute latency in µs for drive API storage operations",
+		append(allDriveLabels, apiL)...)
+
+	driveOfflineCountMD = NewGaugeMD(driveOfflineCount,
+		"Count of offline drives")
+	driveOnlineCountMD = NewGaugeMD(driveOnlineCount,
+		"Count of online drives")
+	driveCountMD = NewGaugeMD(driveCount,
+		"Count of all drives")
+)
+
+// loadDriveMetrics - `MetricsLoaderFn` for node drive metrics.
+func loadDriveMetrics(ctx context.Context, m MetricValues, c *metricsCache) error {
+	driveMetrics, err := c.driveMetrics.Get()
+	if err != nil {
+		logger.LogIf(ctx, err)
+		return nil
+	}
+
+	storageInfo := driveMetrics.storageInfo
+
+	for _, disk := range storageInfo.Disks {
+		labels := []string{
+			driveL, disk.DrivePath,
+			poolIndexL, strconv.Itoa(disk.PoolIndex),
+			setIndexL, strconv.Itoa(disk.SetIndex),
+			driveIndexL, strconv.Itoa(disk.DiskIndex),
+		}
+
+		m.Set(driveUsedBytes, float64(disk.UsedSpace), labels...)
+		m.Set(driveFreeBytes, float64(disk.AvailableSpace), labels...)
+		m.Set(driveTotalBytes, float64(disk.TotalSpace), labels...)
+		m.Set(driveFreeInodes, float64(disk.FreeInodes), labels...)
+
+		if disk.Metrics != nil {
+			m.Set(driveTimeoutErrorsTotal, float64(disk.Metrics.TotalErrorsTimeout), labels...)
+			m.Set(driveAvailabilityErrorsTotal, float64(disk.Metrics.TotalErrorsAvailability), labels...)
+			m.Set(driveWaitingIO, float64(disk.Metrics.TotalWaiting), labels...)
+
+			// Append the api label for the drive API latencies.
+			labels = append(labels, "api", "")
+			lastIdx := len(labels) - 1
+			for apiName, latency := range disk.Metrics.LastMinute {
+				labels[lastIdx] = "storage." + apiName
+				m.Set(driveAPILatencyMicros, float64(latency.Avg().Microseconds()),
+					labels...)
+			}
+		}
+	}
+
+	m.Set(driveOfflineCount, float64(driveMetrics.offlineDrives))
+	m.Set(driveOnlineCount, float64(driveMetrics.onlineDrives))
+	m.Set(driveCount, float64(driveMetrics.totalDrives))
+
+	return nil
+}

cmd/metrics-v3-system-network.go

@@ -0,0 +1,61 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+
+	"github.com/minio/minio/internal/rest"
+)
+
+const (
+	internodeErrorsTotal      MetricName = "errors_total"
+	internodeDialErrorsTotal  MetricName = "dial_errors_total"
+	internodeDialAvgTimeNanos MetricName = "dial_avg_time_nanos"
+	internodeSentBytesTotal   MetricName = "sent_bytes_total"
+	internodeRecvBytesTotal   MetricName = "recv_bytes_total"
+)
+
+var (
+	internodeErrorsTotalMD = NewCounterMD(internodeErrorsTotal,
+		"Total number of failed internode calls")
+	internodeDialedErrorsTotalMD = NewCounterMD(internodeDialErrorsTotal,
+		"Total number of internode TCP dial timeouts and errors")
+	internodeDialAvgTimeNanosMD = NewGaugeMD(internodeDialAvgTimeNanos,
+		"Average dial time of internode TCP calls in nanoseconds")
+	internodeSentBytesTotalMD = NewCounterMD(internodeSentBytesTotal,
+		"Total number of bytes sent to other peer nodes")
+	internodeRecvBytesTotalMD = NewCounterMD(internodeRecvBytesTotal,
+		"Total number of bytes received from other peer nodes")
+)
+
+// loadNetworkInternodeMetrics - reads internode network metrics.
+//
+// This is a `MetricsLoaderFn`.
+func loadNetworkInternodeMetrics(ctx context.Context, m MetricValues, _ *metricsCache) error {
+	connStats := globalConnStats.toServerConnStats()
+	rpcStats := rest.GetRPCStats()
+	if globalIsDistErasure {
+		m.Set(internodeErrorsTotal, float64(rpcStats.Errs))
+		m.Set(internodeDialErrorsTotal, float64(rpcStats.DialErrs))
+		m.Set(internodeDialAvgTimeNanos, float64(rpcStats.DialAvgDuration))
+		m.Set(internodeSentBytesTotal, float64(connStats.internodeOutputBytes))
+		m.Set(internodeRecvBytesTotal, float64(connStats.internodeInputBytes))
+	}
+	return nil
+}

cmd/metrics-v3-types.go

@@ -0,0 +1,487 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"sync"
+
+	"github.com/minio/minio/internal/logger"
+	"github.com/pkg/errors"
+	"github.com/prometheus/client_golang/prometheus"
+	"golang.org/x/exp/slices"
+)
+
+type collectorPath string
+
+// metricPrefix converts a collector path to a metric name prefix. The path is
+// converted to snake-case (by replaced '/' and '-' with '_') and prefixed with
+// `minio_`.
+func (cp collectorPath) metricPrefix() string {
+	s := strings.TrimPrefix(string(cp), "/")
+	s = strings.ReplaceAll(s, "/", "_")
+	s = strings.ReplaceAll(s, "-", "_")
+	return "minio_" + s
+}
+
+// isDescendantOf returns true if it is a descendant of (or the same as)
+// `ancestor`.
+//
+// For example:
+//
+//	 	/a, /a/b, /a/b/c are all descendants of /a.
+//		/abc or /abd/a are not descendants of /ab.
+func (cp collectorPath) isDescendantOf(arg string) bool {
+	descendant := string(cp)
+	if descendant == arg {
+		return true
+	}
+	if len(arg) >= len(descendant) {
+		return false
+	}
+	if !strings.HasSuffix(arg, "/") {
+		arg += "/"
+	}
+	return strings.HasPrefix(descendant, arg)
+}
+
+// MetricType - represents the type of a metric.
+type MetricType int
+
+const (
+	// CounterMT - represents a counter metric.
+	CounterMT MetricType = iota
+	// GaugeMT - represents a gauge metric.
+	GaugeMT
+	// HistogramMT - represents a histogram metric.
+	HistogramMT
+)
+
+func (mt MetricType) String() string {
+	switch mt {
+	case CounterMT:
+		return "counter"
+	case GaugeMT:
+		return "gauge"
+	case HistogramMT:
+		return "histogram"
+	default:
+		return "*unknown*"
+	}
+}
+
+func (mt MetricType) toProm() prometheus.ValueType {
+	switch mt {
+	case CounterMT:
+		return prometheus.CounterValue
+	case GaugeMT:
+		return prometheus.GaugeValue
+	case HistogramMT:
+		return prometheus.CounterValue
+	default:
+		panic(fmt.Sprintf("unknown metric type: %d", mt))
+	}
+}
+
+// MetricDescriptor - represents a metric descriptor.
+type MetricDescriptor struct {
+	Name           MetricName
+	Type           MetricType
+	Help           string
+	VariableLabels []string
+
+	// managed values follow:
+	labelSet map[string]struct{}
+}
+
+func (md *MetricDescriptor) getLabelSet() map[string]struct{} {
+	if md.labelSet != nil {
+		return md.labelSet
+	}
+	md.labelSet = make(map[string]struct{}, len(md.VariableLabels))
+	for _, label := range md.VariableLabels {
+		md.labelSet[label] = struct{}{}
+	}
+	return md.labelSet
+}
+
+func (md *MetricDescriptor) toPromName(namePrefix string) string {
+	return prometheus.BuildFQName(namePrefix, "", string(md.Name))
+}
+
+func (md *MetricDescriptor) toPromDesc(namePrefix string, extraLabels map[string]string) *prometheus.Desc {
+	return prometheus.NewDesc(
+		md.toPromName(namePrefix),
+		md.Help,
+		md.VariableLabels, extraLabels,
+	)
+}
+
+// NewCounterMD - creates a new counter metric descriptor.
+func NewCounterMD(name MetricName, help string, labels ...string) MetricDescriptor {
+	return MetricDescriptor{
+		Name:           name,
+		Type:           CounterMT,
+		Help:           help,
+		VariableLabels: labels,
+	}
+}
+
+// NewGaugeMD - creates a new gauge metric descriptor.
+func NewGaugeMD(name MetricName, help string, labels ...string) MetricDescriptor {
+	return MetricDescriptor{
+		Name:           name,
+		Type:           GaugeMT,
+		Help:           help,
+		VariableLabels: labels,
+	}
+}
+
+type metricValue struct {
+	Labels map[string]string
+	Value  float64
+}
+
+// MetricValues - type to set metric values retrieved while loading metrics. A
+// value of this type is passed to the `MetricsLoaderFn`.
+type MetricValues struct {
+	values      map[MetricName][]metricValue
+	descriptors map[MetricName]MetricDescriptor
+}
+
+func newMetricValues(d map[MetricName]MetricDescriptor) MetricValues {
+	return MetricValues{
+		values:      make(map[MetricName][]metricValue, len(d)),
+		descriptors: d,
+	}
+}
+
+// ToPromMetrics - converts the internal metric values to Prometheus
+// adding the given name prefix. The extraLabels are added to each metric as
+// constant labels.
+func (m *MetricValues) ToPromMetrics(namePrefix string, extraLabels map[string]string,
+) []prometheus.Metric {
+	metrics := make([]prometheus.Metric, 0, len(m.values))
+	for metricName, mv := range m.values {
+		desc := m.descriptors[metricName]
+		promDesc := desc.toPromDesc(namePrefix, extraLabels)
+		for _, v := range mv {
+			// labelValues is in the same order as the variable labels in the
+			// descriptor.
+			labelValues := make([]string, 0, len(v.Labels))
+			for _, k := range desc.VariableLabels {
+				labelValues = append(labelValues, v.Labels[k])
+			}
+			metrics = append(metrics,
+				prometheus.MustNewConstMetric(promDesc, desc.Type.toProm(), v.Value,
+					labelValues...))
+		}
+	}
+	return metrics
+}
+
+// Set - sets a metric value along with any provided labels. It is used only
+// with Gauge and Counter metrics.
+//
+// If the MetricName given here is not present in the `MetricsGroup`'s
+// descriptors, this function panics.
+//
+// Panics if `labels` is not a list of ordered label name and label value pairs
+// or if all labels for the metric are not provided.
+func (m *MetricValues) Set(name MetricName, value float64, labels ...string) {
+	desc, ok := m.descriptors[name]
+	if !ok {
+		panic(fmt.Sprintf("metric has no description: %s", name))
+	}
+
+	if len(labels)%2 != 0 {
+		panic("labels must be a list of ordered key-value pairs")
+	}
+
+	validLabels := desc.getLabelSet()
+	labelMap := make(map[string]string, len(labels)/2)
+	for i := 0; i < len(labels); i += 2 {
+		if _, ok := validLabels[labels[i]]; !ok {
+			panic(fmt.Sprintf("invalid label: %s (metric: %s)", labels[i], name))
+		}
+		labelMap[labels[i]] = labels[i+1]
+	}
+
+	if len(labels)/2 != len(validLabels) {
+		panic(fmt.Sprintf("not all labels were given values"))
+	}
+
+	v, ok := m.values[name]
+	if !ok {
+		v = make([]metricValue, 0, 1)
+	}
+	m.values[name] = append(v, metricValue{
+		Labels: labelMap,
+		Value:  value,
+	})
+}
+
+// SetHistogram - sets values for the given MetricName using the provided
+// histogram.
+//
+// `renameLabels` is a map of label names to rename. The keys are the original
+// label names and the values are the new label names.
+//
+// TODO: bucketFilter doc
+//
+// `extraLabels` are additional labels to add to each metric. They are ordered
+// label name and value pairs.
+func (m *MetricValues) SetHistogram(name MetricName, hist *prometheus.HistogramVec,
+	renameLabels map[string]string, bucketFilter []string, extraLabels ...string,
+) {
+	if _, ok := m.descriptors[name]; !ok {
+		panic(fmt.Sprintf("metric has no description: %s", name))
+	}
+	dummyDesc := MetricDescription{}
+	metricsV2 := getHistogramMetrics(hist, dummyDesc, false)
+	for _, metric := range metricsV2 {
+		// If a bucket filter is provided, only add metrics for the given
+		// buckets.
+		if len(bucketFilter) > 0 {
+			if !slices.Contains(bucketFilter, metric.VariableLabels["bucket"]) {
+				continue
+			}
+		}
+
+		labels := make([]string, 0, len(metric.VariableLabels)*2)
+		for k, v := range metric.VariableLabels {
+			if newLabel, ok := renameLabels[k]; ok {
+				labels = append(labels, newLabel, v)
+			} else {
+				labels = append(labels, k, v)
+			}
+		}
+		labels = append(labels, extraLabels...)
+		m.Set(name, metric.Value, labels...)
+	}
+}
+
+// MetricsLoaderFn - represents a function to load metrics from the
+// metricsCache.
+//
+// Note that returning an error here will cause the Metrics handler to return a
+// 500 Internal Server Error.
+type MetricsLoaderFn func(context.Context, MetricValues, *metricsCache) error
+
+// JoinLoaders - joins multiple loaders into a single loader. The returned
+// loader will call each of the given loaders in order. If any of the loaders
+// return an error, the returned loader will return that error.
+func JoinLoaders(loaders ...MetricsLoaderFn) MetricsLoaderFn {
+	return func(ctx context.Context, m MetricValues, c *metricsCache) error {
+		for _, loader := range loaders {
+			if err := loader(ctx, m, c); err != nil {
+				return err
+			}
+		}
+		return nil
+	}
+}
+
+// BucketMetricsLoaderFn - represents a function to load metrics from the
+// metricsCache and the system for a given list of buckets.
+//
+// Note that returning an error here will cause the Metrics handler to return a
+// 500 Internal Server Error.
+type BucketMetricsLoaderFn func(context.Context, MetricValues, *metricsCache, []string) error
+
+// JoinBucketLoaders - joins multiple bucket loaders into a single loader,
+// similar to `JoinLoaders`.
+func JoinBucketLoaders(loaders ...BucketMetricsLoaderFn) BucketMetricsLoaderFn {
+	return func(ctx context.Context, m MetricValues, c *metricsCache, b []string) error {
+		for _, loader := range loaders {
+			if err := loader(ctx, m, c, b); err != nil {
+				return err
+			}
+		}
+		return nil
+	}
+}
+
+// MetricsGroup - represents a group of metrics. It includes a `MetricsLoaderFn`
+// function that provides a way to load the metrics from the system. The metrics
+// are cached and refreshed after a given timeout.
+//
+// For metrics with a `bucket` dimension, a list of buckets argument is required
+// to collect the metrics.
+//
+// It implements the prometheus.Collector interface for metric groups without a
+// bucket dimension. For metric groups with a bucket dimension, use the
+// `GetBucketCollector` method to get a `BucketCollector` that implements the
+// prometheus.Collector interface.
+type MetricsGroup struct {
+	// Path (relative to the Metrics v3 base endpoint) at which this group of
+	// metrics is served. This value is converted into a metric name prefix
+	// using `.metricPrefix()` and is added to each metric returned.
+	CollectorPath collectorPath
+	// List of all metric descriptors that could be returned by the loader.
+	Descriptors []MetricDescriptor
+	// (Optional) Extra (constant) label KV pairs to be added to each metric in
+	// the group.
+	ExtraLabels map[string]string
+
+	// Loader functions to load metrics. Only one of these will be set. Metrics
+	// returned by these functions must be present in the `Descriptors` list.
+	loader       MetricsLoaderFn
+	bucketLoader BucketMetricsLoaderFn
+
+	// Cache for all metrics groups. Set via `.SetCache` method.
+	cache *metricsCache
+
+	// managed values follow:
+
+	// map of metric descriptors by metric name.
+	descriptorMap map[MetricName]MetricDescriptor
+
+	// For bucket metrics, the list of buckets is stored here. It is used in the
+	// Collect() call. This is protected by the `bucketsLock`.
+	bucketsLock sync.Mutex
+	buckets     []string
+}
+
+// NewMetricsGroup creates a new MetricsGroup. To create a metrics group for
+// metrics with a `bucket` dimension (label), use `NewBucketMetricsGroup`.
+//
+// The `loader` function loads metrics from the cache and the system.
+func NewMetricsGroup(path collectorPath, descriptors []MetricDescriptor,
+	loader MetricsLoaderFn,
+) *MetricsGroup {
+	mg := &MetricsGroup{
+		CollectorPath: path,
+		Descriptors:   descriptors,
+		loader:        loader,
+	}
+	mg.validate()
+	return mg
+}
+
+// NewBucketMetricsGroup creates a new MetricsGroup for metrics with a `bucket`
+// dimension (label).
+//
+// The `loader` function loads metrics from the cache and the system for a given
+// list of buckets.
+func NewBucketMetricsGroup(path collectorPath, descriptors []MetricDescriptor,
+	loader BucketMetricsLoaderFn,
+) *MetricsGroup {
+	mg := &MetricsGroup{
+		CollectorPath: path,
+		Descriptors:   descriptors,
+		bucketLoader:  loader,
+	}
+	mg.validate()
+	return mg
+}
+
+// AddExtraLabels - adds extra (constant) label KV pairs to the metrics group.
+// This is a helper to initialize the `ExtraLabels` field. The argument is a
+// list of ordered label name and value pairs.
+func (mg *MetricsGroup) AddExtraLabels(labels ...string) {
+	if len(labels)%2 != 0 {
+		panic("Labels must be an ordered list of name value pairs")
+	}
+	if mg.ExtraLabels == nil {
+		mg.ExtraLabels = make(map[string]string, len(labels))
+	}
+	for i := 0; i < len(labels); i += 2 {
+		mg.ExtraLabels[labels[i]] = labels[i+1]
+	}
+}
+
+// IsBucketMetricsGroup - returns true if the given MetricsGroup is a bucket
+// metrics group.
+func (mg *MetricsGroup) IsBucketMetricsGroup() bool {
+	return mg.bucketLoader != nil
+}
+
+// Describe - implements prometheus.Collector interface.
+func (mg *MetricsGroup) Describe(ch chan<- *prometheus.Desc) {
+	for _, desc := range mg.Descriptors {
+		ch <- desc.toPromDesc(mg.CollectorPath.metricPrefix(), mg.ExtraLabels)
+	}
+}
+
+// Collect - implements prometheus.Collector interface.
+func (mg *MetricsGroup) Collect(ch chan<- prometheus.Metric) {
+	metricValues := newMetricValues(mg.descriptorMap)
+
+	var err error
+	if mg.IsBucketMetricsGroup() {
+		err = mg.bucketLoader(GlobalContext, metricValues, mg.cache, mg.buckets)
+	} else {
+		err = mg.loader(GlobalContext, metricValues, mg.cache)
+	}
+
+	// There is no way to handle errors here, so we panic the current goroutine
+	// and the Metrics API handler returns a 500 HTTP status code. This should
+	// normally not happen, and usually indicates a bug.
+	logger.CriticalIf(GlobalContext, errors.Wrap(err, "failed to get metrics"))
+
+	promMetrics := metricValues.ToPromMetrics(mg.CollectorPath.metricPrefix(),
+		mg.ExtraLabels)
+	for _, metric := range promMetrics {
+		ch <- metric
+	}
+}
+
+// LockAndSetBuckets - locks the buckets and sets the given buckets. It returns
+// a function to unlock the buckets.
+func (mg *MetricsGroup) LockAndSetBuckets(buckets []string) func() {
+	mg.bucketsLock.Lock()
+	mg.buckets = buckets
+	return func() {
+		mg.bucketsLock.Unlock()
+	}
+}
+
+// MetricFQN - returns the fully qualified name for the given metric name.
+func (mg *MetricsGroup) MetricFQN(name MetricName) string {
+	v, ok := mg.descriptorMap[name]
+	if !ok {
+		// This should never happen.
+		return ""
+	}
+	return v.toPromName(mg.CollectorPath.metricPrefix())
+}
+
+func (mg *MetricsGroup) validate() {
+	if len(mg.Descriptors) == 0 {
+		panic("Descriptors must be set")
+	}
+
+	// For bools A and B, A XOR B <=> A != B.
+	isExactlyOneSet := (mg.loader == nil) != (mg.bucketLoader == nil)
+	if !isExactlyOneSet {
+		panic("Exactly one Loader function must be set")
+	}
+
+	mg.descriptorMap = make(map[MetricName]MetricDescriptor, len(mg.Descriptors))
+	for _, desc := range mg.Descriptors {
+		mg.descriptorMap[desc.Name] = desc
+	}
+}
+
+// SetCache is a helper to initialize MetricsGroup. It sets the cache object.
+func (mg *MetricsGroup) SetCache(c *metricsCache) {
+	mg.cache = c
+}

cmd/metrics-v3.go

@@ -0,0 +1,272 @@
+// Copyright (c) 2015-2024 MinIO, Inc.
+//
+// This file is part of MinIO Object Storage stack
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package cmd
+
+import (
+	"slices"
+	"strings"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/collectors"
+)
+
+// Collector paths.
+//
+// These are paths under the top-level /minio/metrics/v3 metrics endpoint. Each
+// of these paths returns a set of V3 metrics.
+const (
+	apiRequestsCollectorPath collectorPath = "/api/requests"
+	apiBucketCollectorPath   collectorPath = "/api/bucket"
+
+	systemNetworkInternodeCollectorPath collectorPath = "/system/network/internode"
+	systemDriveCollectorPath            collectorPath = "/system/drive"
+	systemProcessCollectorPath          collectorPath = "/system/process"
+	systemGoCollectorPath               collectorPath = "/system/go"
+
+	clusterHealthCollectorPath       collectorPath = "/cluster/health"
+	clusterUsageObjectsCollectorPath collectorPath = "/cluster/usage/objects"
+	clusterUsageBucketsCollectorPath collectorPath = "/cluster/usage/buckets"
+	clusterErasureSetCollectorPath   collectorPath = "/cluster/erasure-set"
+)
+
+const (
+	clusterBasePath = "/cluster"
+)
+
+type metricsV3Collection struct {
+	mgMap       map[collectorPath]*MetricsGroup
+	bucketMGMap map[collectorPath]*MetricsGroup
+
+	// Gatherers for non-bucket MetricsGroup's
+	mgGatherers map[collectorPath]prometheus.Gatherer
+
+	collectorPaths []collectorPath
+}
+
+func newMetricGroups(r *prometheus.Registry) *metricsV3Collection {
+	// Create all metric groups.
+	apiRequestsMG := NewMetricsGroup(apiRequestsCollectorPath,
+		[]MetricDescriptor{
+			apiRejectedAuthTotalMD,
+			apiRejectedHeaderTotalMD,
+			apiRejectedTimestampTotalMD,
+			apiRejectedInvalidTotalMD,
+
+			apiRequestsWaitingTotalMD,
+			apiRequestsIncomingTotalMD,
+			apiRequestsInFlightTotalMD,
+			apiRequestsTotalMD,
+			apiRequestsErrorsTotalMD,
+			apiRequests5xxErrorsTotalMD,
+			apiRequests4xxErrorsTotalMD,
+			apiRequestsCanceledTotalMD,
+
+			apiRequestsTTFBSecondsDistributionMD,
+
+			apiTrafficSentBytesMD,
+			apiTrafficRecvBytesMD,
+		},
+		JoinLoaders(loadAPIRequestsHTTPMetrics, loadAPIRequestsTTFBMetrics,
+			loadAPIRequestsNetworkMetrics),
+	)
+
+	apiBucketMG := NewBucketMetricsGroup(apiBucketCollectorPath,
+		[]MetricDescriptor{
+			apiBucketTrafficRecvBytesMD,
+			apiBucketTrafficSentBytesMD,
+
+			apiBucketRequestsInFlightMD,
+			apiBucketRequestsTotalMD,
+			apiBucketRequestsCanceledMD,
+			apiBucketRequests4xxErrorsMD,
+			apiBucketRequests5xxErrorsMD,
+
+			apiBucketRequestsTTFBSecondsDistributionMD,
+		},
+		JoinBucketLoaders(loadAPIBucketHTTPMetrics, loadAPIBucketTTFBMetrics),
+	)
+
+	systemNetworkInternodeMG := NewMetricsGroup(systemNetworkInternodeCollectorPath,
+		[]MetricDescriptor{
+			internodeErrorsTotalMD,
+			internodeDialedErrorsTotalMD,
+			internodeDialAvgTimeNanosMD,
+			internodeSentBytesTotalMD,
+			internodeRecvBytesTotalMD,
+		},
+		loadNetworkInternodeMetrics,
+	)
+
+	systemDriveMG := NewMetricsGroup(systemDriveCollectorPath,
+		[]MetricDescriptor{
+			driveUsedBytesMD,
+			driveFreeBytesMD,
+			driveTotalBytesMD,
+			driveFreeInodesMD,
+			driveTimeoutErrorsMD,
+			driveAvailabilityErrorsMD,
+			driveWaitingIOMD,
+			driveAPILatencyMD,
+
+			driveOfflineCountMD,
+			driveOnlineCountMD,
+			driveCountMD,
+		},
+		loadDriveMetrics,
+	)
+
+	clusterHealthMG := NewMetricsGroup(clusterHealthCollectorPath,
+		[]MetricDescriptor{
+			healthDrivesOfflineCountMD,
+			healthDrivesOnlineCountMD,
+			healthDrivesCountMD,
+
+			healthNodesOfflineCountMD,
+			healthNodesOnlineCountMD,
+
+			healthCapacityRawTotalBytesMD,
+			healthCapacityRawFreeBytesMD,
+			healthCapacityUsableTotalBytesMD,
+			healthCapacityUsableFreeBytesMD,
+		},
+		JoinLoaders(loadClusterHealthDriveMetrics,
+			loadClusterHealthNodeMetrics,
+			loadClusterHealthCapacityMetrics),
+	)
+
+	clusterUsageObjectsMG := NewMetricsGroup(clusterUsageObjectsCollectorPath,
+		[]MetricDescriptor{
+			usageSinceLastUpdateSecondsMD,
+			usageTotalBytesMD,
+			usageObjectsCountMD,
+			usageVersionsCountMD,
+			usageDeleteMarkersCountMD,
+			usageBucketsCountMD,
+			usageObjectsDistributionMD,
+			usageVersionsDistributionMD,
+		},
+		loadClusterUsageObjectMetrics,
+	)
+
+	clusterUsageBucketsMG := NewBucketMetricsGroup(clusterUsageBucketsCollectorPath,
+		[]MetricDescriptor{
+			usageSinceLastUpdateSecondsMD,
+			usageBucketTotalBytesMD,
+			usageBucketObjectsTotalMD,
+			usageBucketVersionsCountMD,
+			usageBucketDeleteMarkersCountMD,
+			usageBucketQuotaTotalBytesMD,
+			usageBucketObjectSizeDistributionMD,
+			usageBucketObjectVersionCountDistributionMD,
+		},
+		loadClusterUsageBucketMetrics,
+	)
+
+	clusterErasureSetMG := NewMetricsGroup(clusterErasureSetCollectorPath,
+		[]MetricDescriptor{
+			erasureSetOverallWriteQuorumMD,
+			erasureSetOverallHealthMD,
+			erasureSetReadQuorumMD,
+			erasureSetWriteQuorumMD,
+			erasureSetOnlineDrivesCountMD,
+			erasureSetHealingDrivesCountMD,
+			erasureSetHealthMD,
+		},
+		loadClusterErasureSetMetrics,
+	)
+
+	allMetricGroups := []*MetricsGroup{
+		apiRequestsMG,
+		apiBucketMG,
+
+		systemNetworkInternodeMG,
+		systemDriveMG,
+
+		clusterHealthMG,
+		clusterUsageObjectsMG,
+		clusterUsageBucketsMG,
+		clusterErasureSetMG,
+	}
+
+	// Bucket metrics are special, they always include the bucket label. These
+	// metrics required a list of buckets to be passed to the loader, and the list
+	// of buckets is not known until the request is made. So we keep a separate
+	// map for bucket metrics and handle them specially.
+
+	// Add the serverName and poolIndex labels to all non-cluster metrics.
+	//
+	// Also create metric group maps and set the cache.
+	metricsCache := newMetricsCache()
+	mgMap := make(map[collectorPath]*MetricsGroup)
+	bucketMGMap := make(map[collectorPath]*MetricsGroup)
+	for _, mg := range allMetricGroups {
+		if !strings.HasPrefix(string(mg.CollectorPath), clusterBasePath) {
+			mg.AddExtraLabels(
+				serverName, globalLocalNodeName,
+				// poolIndex, strconv.Itoa(globalLocalPoolIdx),
+			)
+		}
+		mg.SetCache(metricsCache)
+		if mg.IsBucketMetricsGroup() {
+			bucketMGMap[mg.CollectorPath] = mg
+		} else {
+			mgMap[mg.CollectorPath] = mg
+		}
+	}
+
+	// Prepare to register the collectors. Other than `MetricGroup` collectors,
+	// we also have standard collectors like `ProcessCollector` and `GoCollector`.
+
+	// Create all Non-`MetricGroup` collectors here.
+	collectors := map[collectorPath]prometheus.Collector{
+		systemProcessCollectorPath: collectors.NewProcessCollector(collectors.ProcessCollectorOpts{
+			ReportErrors: true,
+		}),
+		systemGoCollectorPath: collectors.NewGoCollector(),
+	}
+
+	// Add all `MetricGroup` collectors to the map.
+	for _, mg := range allMetricGroups {
+		collectors[mg.CollectorPath] = mg
+	}
+
+	// Helper function to register a collector and return a gatherer for it.
+	mustRegister := func(c ...prometheus.Collector) prometheus.Gatherer {
+		subRegistry := prometheus.NewRegistry()
+		for _, col := range c {
+			subRegistry.MustRegister(col)
+		}
+		r.MustRegister(subRegistry)
+		return subRegistry
+	}
+
+	// Register all collectors and create gatherers for them.
+	gatherers := make(map[collectorPath]prometheus.Gatherer, len(collectors))
+	collectorPaths := make([]collectorPath, 0, len(collectors))
+	for path, collector := range collectors {
+		gatherers[path] = mustRegister(collector)
+		collectorPaths = append(collectorPaths, path)
+	}
+	slices.Sort(collectorPaths)
+	return &metricsV3Collection{
+		mgMap:          mgMap,
+		bucketMGMap:    bucketMGMap,
+		mgGatherers:    gatherers,
+		collectorPaths: collectorPaths,
+	}
+}

cmd/notification.go

@@ -870,12 +870,12 @@ func (sys *NotificationSys) GetMetrics(ctx context.Context, t madmin.MetricType,
 }
 
 // GetResourceMetrics - gets the resource metrics from all nodes excluding self.
-func (sys *NotificationSys) GetResourceMetrics(ctx context.Context) <-chan Metric {
+func (sys *NotificationSys) GetResourceMetrics(ctx context.Context) <-chan MetricV2 {
 	if sys == nil {
 		return nil
 	}
 	g := errgroup.WithNErrs(len(sys.peerClients))
-	peerChannels := make([]<-chan Metric, len(sys.peerClients))
+	peerChannels := make([]<-chan MetricV2, len(sys.peerClients))
 	for index := range sys.peerClients {
 		index := index
 		g.Go(func() error {
@@ -1214,8 +1214,8 @@ func (sys *NotificationSys) GetBandwidthReports(ctx context.Context, buckets ...
 	return consolidatedReport
 }
 
-func (sys *NotificationSys) collectPeerMetrics(ctx context.Context, peerChannels []<-chan Metric, g *errgroup.Group) <-chan Metric {
-	ch := make(chan Metric)
+func (sys *NotificationSys) collectPeerMetrics(ctx context.Context, peerChannels []<-chan MetricV2, g *errgroup.Group) <-chan MetricV2 {
+	ch := make(chan MetricV2)
 	var wg sync.WaitGroup
 	for index, err := range g.Wait() {
 		if err != nil {
@@ -1229,7 +1229,7 @@ func (sys *NotificationSys) collectPeerMetrics(ctx context.Context, peerChannels
 			continue
 		}
 		wg.Add(1)
-		go func(ctx context.Context, peerChannel <-chan Metric, wg *sync.WaitGroup) {
+		go func(ctx context.Context, peerChannel <-chan MetricV2, wg *sync.WaitGroup) {
 			defer wg.Done()
 			for {
 				select {
@@ -1248,7 +1248,7 @@ func (sys *NotificationSys) collectPeerMetrics(ctx context.Context, peerChannels
 			}
 		}(ctx, peerChannels[index], &wg)
 	}
-	go func(wg *sync.WaitGroup, ch chan Metric) {
+	go func(wg *sync.WaitGroup, ch chan MetricV2) {
 		wg.Wait()
 		xioutil.SafeClose(ch)
 	}(&wg, ch)
@@ -1256,12 +1256,12 @@ func (sys *NotificationSys) collectPeerMetrics(ctx context.Context, peerChannels
 }
 
 // GetBucketMetrics - gets the cluster level bucket metrics from all nodes excluding self.
-func (sys *NotificationSys) GetBucketMetrics(ctx context.Context) <-chan Metric {
+func (sys *NotificationSys) GetBucketMetrics(ctx context.Context) <-chan MetricV2 {
 	if sys == nil {
 		return nil
 	}
 	g := errgroup.WithNErrs(len(sys.peerClients))
-	peerChannels := make([]<-chan Metric, len(sys.peerClients))
+	peerChannels := make([]<-chan MetricV2, len(sys.peerClients))
 	for index := range sys.peerClients {
 		index := index
 		g.Go(func() error {
@@ -1277,12 +1277,12 @@ func (sys *NotificationSys) GetBucketMetrics(ctx context.Context) <-chan Metric
 }
 
 // GetClusterMetrics - gets the cluster metrics from all nodes excluding self.
-func (sys *NotificationSys) GetClusterMetrics(ctx context.Context) <-chan Metric {
+func (sys *NotificationSys) GetClusterMetrics(ctx context.Context) <-chan MetricV2 {
 	if sys == nil {
 		return nil
 	}
 	g := errgroup.WithNErrs(len(sys.peerClients))
-	peerChannels := make([]<-chan Metric, len(sys.peerClients))
+	peerChannels := make([]<-chan MetricV2, len(sys.peerClients))
 	for index := range sys.peerClients {
 		index := index
 		g.Go(func() error {

cmd/peer-rest-client.go

@@ -641,13 +641,13 @@ func (client *peerRESTClient) MonitorBandwidth(ctx context.Context, buckets []st
 	return getBandwidthRPC.Call(ctx, client.gridConn(), values)
 }
 
-func (client *peerRESTClient) GetResourceMetrics(ctx context.Context) (<-chan Metric, error) {
+func (client *peerRESTClient) GetResourceMetrics(ctx context.Context) (<-chan MetricV2, error) {
 	resp, err := getResourceMetricsRPC.Call(ctx, client.gridConn(), grid.NewMSS())
 	if err != nil {
 		return nil, err
 	}
-	ch := make(chan Metric)
-	go func(ch chan<- Metric) {
+	ch := make(chan MetricV2)
+	go func(ch chan<- MetricV2) {
 		defer close(ch)
 		for _, m := range resp.Value() {
 			if m == nil {
@@ -663,12 +663,12 @@ func (client *peerRESTClient) GetResourceMetrics(ctx context.Context) (<-chan Me
 	return ch, nil
 }
 
-func (client *peerRESTClient) GetPeerMetrics(ctx context.Context) (<-chan Metric, error) {
+func (client *peerRESTClient) GetPeerMetrics(ctx context.Context) (<-chan MetricV2, error) {
 	resp, err := getPeerMetricsRPC.Call(ctx, client.gridConn(), grid.NewMSS())
 	if err != nil {
 		return nil, err
 	}
-	ch := make(chan Metric)
+	ch := make(chan MetricV2)
 	go func() {
 		defer close(ch)
 		for _, m := range resp.Value() {
@@ -685,12 +685,12 @@ func (client *peerRESTClient) GetPeerMetrics(ctx context.Context) (<-chan Metric
 	return ch, nil
 }
 
-func (client *peerRESTClient) GetPeerBucketMetrics(ctx context.Context) (<-chan Metric, error) {
+func (client *peerRESTClient) GetPeerBucketMetrics(ctx context.Context) (<-chan MetricV2, error) {
 	resp, err := getPeerBucketMetricsRPC.Call(ctx, client.gridConn(), grid.NewMSS())
 	if err != nil {
 		return nil, err
 	}
-	ch := make(chan Metric)
+	ch := make(chan MetricV2)
 	go func() {
 		defer close(ch)
 		for _, m := range resp.Value() {

cmd/peer-rest-server.go

@@ -54,7 +54,7 @@ type peerRESTServer struct{}
 var (
 	// Types & Wrappers
 	aoBucketInfo           = grid.NewArrayOf[*BucketInfo](func() *BucketInfo { return &BucketInfo{} })
-	aoMetricsGroup         = grid.NewArrayOf[*Metric](func() *Metric { return &Metric{} })
+	aoMetricsGroup         = grid.NewArrayOf[*MetricV2](func() *MetricV2 { return &MetricV2{} })
 	madminBgHealState      = grid.NewJSONPool[madmin.BgHealState]()
 	madminCPUs             = grid.NewJSONPool[madmin.CPUs]()
 	madminMemInfo          = grid.NewJSONPool[madmin.MemInfo]()
@@ -88,9 +88,9 @@ var (
 	getNetInfoRPC                  = grid.NewSingleHandler[*grid.MSS, *grid.JSON[madmin.NetInfo]](grid.HandlerGetNetInfo, grid.NewMSS, madminNetInfo.NewJSON)
 	getOSInfoRPC                   = grid.NewSingleHandler[*grid.MSS, *grid.JSON[madmin.OSInfo]](grid.HandlerGetOSInfo, grid.NewMSS, madminOSInfo.NewJSON)
 	getPartitionsRPC               = grid.NewSingleHandler[*grid.MSS, *grid.JSON[madmin.Partitions]](grid.HandlerGetPartitions, grid.NewMSS, madminPartitions.NewJSON)
-	getPeerBucketMetricsRPC        = grid.NewSingleHandler[*grid.MSS, *grid.Array[*Metric]](grid.HandlerGetPeerBucketMetrics, grid.NewMSS, aoMetricsGroup.New)
-	getPeerMetricsRPC              = grid.NewSingleHandler[*grid.MSS, *grid.Array[*Metric]](grid.HandlerGetPeerMetrics, grid.NewMSS, aoMetricsGroup.New)
-	getResourceMetricsRPC          = grid.NewSingleHandler[*grid.MSS, *grid.Array[*Metric]](grid.HandlerGetResourceMetrics, grid.NewMSS, aoMetricsGroup.New)
+	getPeerBucketMetricsRPC        = grid.NewSingleHandler[*grid.MSS, *grid.Array[*MetricV2]](grid.HandlerGetPeerBucketMetrics, grid.NewMSS, aoMetricsGroup.New)
+	getPeerMetricsRPC              = grid.NewSingleHandler[*grid.MSS, *grid.Array[*MetricV2]](grid.HandlerGetPeerMetrics, grid.NewMSS, aoMetricsGroup.New)
+	getResourceMetricsRPC          = grid.NewSingleHandler[*grid.MSS, *grid.Array[*MetricV2]](grid.HandlerGetResourceMetrics, grid.NewMSS, aoMetricsGroup.New)
 	getProcInfoRPC                 = grid.NewSingleHandler[*grid.MSS, *grid.JSON[madmin.ProcInfo]](grid.HandlerGetProcInfo, grid.NewMSS, madminProcInfo.NewJSON)
 	getSRMetricsRPC                = grid.NewSingleHandler[*grid.MSS, *SRMetricsSummary](grid.HandlerGetSRMetrics, grid.NewMSS, func() *SRMetricsSummary { return &SRMetricsSummary{} })
 	getSysConfigRPC                = grid.NewSingleHandler[*grid.MSS, *grid.JSON[madmin.SysConfig]](grid.HandlerGetSysConfig, grid.NewMSS, madminSysConfig.NewJSON)
@@ -1000,9 +1000,9 @@ func (s *peerRESTServer) GetBandwidth(params *grid.URLValues) (*bandwidth.Bucket
 	return globalBucketMonitor.GetReport(selectBuckets), nil
 }
 
-func (s *peerRESTServer) GetResourceMetrics(_ *grid.MSS) (*grid.Array[*Metric], *grid.RemoteErr) {
-	res := make([]*Metric, 0, len(resourceMetricsGroups))
-	populateAndPublish(resourceMetricsGroups, func(m Metric) bool {
+func (s *peerRESTServer) GetResourceMetrics(_ *grid.MSS) (*grid.Array[*MetricV2], *grid.RemoteErr) {
+	res := make([]*MetricV2, 0, len(resourceMetricsGroups))
+	populateAndPublish(resourceMetricsGroups, func(m MetricV2) bool {
 		if m.VariableLabels == nil {
 			m.VariableLabels = make(map[string]string, 1)
 		}
@@ -1014,9 +1014,9 @@ func (s *peerRESTServer) GetResourceMetrics(_ *grid.MSS) (*grid.Array[*Metric],
 }
 
 // GetPeerMetrics gets the metrics to be federated across peers.
-func (s *peerRESTServer) GetPeerMetrics(_ *grid.MSS) (*grid.Array[*Metric], *grid.RemoteErr) {
-	res := make([]*Metric, 0, len(peerMetricsGroups))
-	populateAndPublish(peerMetricsGroups, func(m Metric) bool {
+func (s *peerRESTServer) GetPeerMetrics(_ *grid.MSS) (*grid.Array[*MetricV2], *grid.RemoteErr) {
+	res := make([]*MetricV2, 0, len(peerMetricsGroups))
+	populateAndPublish(peerMetricsGroups, func(m MetricV2) bool {
 		if m.VariableLabels == nil {
 			m.VariableLabels = make(map[string]string, 1)
 		}
@@ -1028,9 +1028,9 @@ func (s *peerRESTServer) GetPeerMetrics(_ *grid.MSS) (*grid.Array[*Metric], *gri
 }
 
 // GetPeerBucketMetrics gets the metrics to be federated across peers.
-func (s *peerRESTServer) GetPeerBucketMetrics(_ *grid.MSS) (*grid.Array[*Metric], *grid.RemoteErr) {
-	res := make([]*Metric, 0, len(bucketPeerMetricsGroups))
-	populateAndPublish(bucketPeerMetricsGroups, func(m Metric) bool {
+func (s *peerRESTServer) GetPeerBucketMetrics(_ *grid.MSS) (*grid.Array[*MetricV2], *grid.RemoteErr) {
+	res := make([]*MetricV2, 0, len(bucketPeerMetricsGroups))
+	populateAndPublish(bucketPeerMetricsGroups, func(m MetricV2) bool {
 		if m.VariableLabels == nil {
 			m.VariableLabels = make(map[string]string, 1)
 		}

cmd/tier.go

@@ -158,17 +158,17 @@ var (
 	}
 )
 
-func (t *tierMetrics) Report() []Metric {
+func (t *tierMetrics) Report() []MetricV2 {
 	metrics := getHistogramMetrics(t.histogram, tierTTLBMD, true)
 	t.RLock()
 	defer t.RUnlock()
 	for tier, stat := range t.requestsCount {
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    tierRequestsSuccessMD,
 			Value:          float64(stat.success),
 			VariableLabels: map[string]string{"tier": tier},
 		})
-		metrics = append(metrics, Metric{
+		metrics = append(metrics, MetricV2{
 			Description:    tierRequestsFailureMD,
 			Value:          float64(stat.failure),
 			VariableLabels: map[string]string{"tier": tier},

docs/metrics/v3.md

@@ -0,0 +1,178 @@
+# Metrics Version 3
+
+In metrics version 3, all metrics are available under the endpoint:
+
+```
+/minio/metrics/v3
+```
+
+however, a specific path under this is required.
+
+Metrics are organized into groups at paths **relative** to the top-level endpoint above.
+
+## Metrics Request Handling
+
+Each endpoint below can be queried at different intervals as needed via a scrape configuration in Prometheus or a compatible metrics collection tool.
+
+For ease of configuration, each (non-empty) parent of the path serves all metric endpoints that are at descendant paths. For example, to query all system metrics one needs to only scrape `/minio/metrics/v3/system/`.
+
+Some metrics are bucket specific. These will have a `/bucket` component in their path. As the number of buckets can be large, the metrics scrape operation needs to be provided with a specific list of buckets via the `bucket` query parameter. Only metrics for the given buckets will be returned (with the bucket label set). For example to query API metrics for buckets `test1` and `test2`, make a scrape request to `/minio/metrics/v3/api/bucket?buckets=test1,test2`. 
+
+Instead of a metrics scrape, it is also possible to list the metrics that would be returned by a path. This is done by adding a `?list` query parameter. The MinIO server will then list all possible metrics that could be returned. During an actual metrics scrape, only available metrics are returned - not all of them. With the `list` query parameter, the output format can be selected - just set the request `Content-Type` to `application/json` for JSON output, or `text/plain` for a simple markdown formatted table. The latter is the default.
+
+## Request, System and Cluster Metrics
+
+At a high level metrics are grouped into three categories, listed in the following sub-sections. The path in each of the tables is relative to the top-level endpoint.
+
+### Request metrics 
+
+These are metrics about requests served by the (current) node.
+
+| Path            | Description                                      |
+|-----------------|--------------------------------------------------|
+| `/api/requests` | Metrics over all requests                        |
+| `/api/bucket`   | Metrics over all requests split by bucket labels |
+|                 |                                                  |
+
+
+### System metrics
+
+These are metrics about the minio process and the node.
+
+| Path                        | Description                                       |
+|-----------------------------|---------------------------------------------------|
+| `/system/drive`             | Metrics about drives on the system                |
+| `/system/network/internode` | Metrics about internode requests made by the node |
+| `/system/process`           | Standard process metrics                          |
+| `/system/go`                | Standard Go lang metrics                          |
+|                             |                                                   |
+
+### Cluster metrics
+
+These present metrics about the whole MinIO cluster.
+
+| Path                     | Description                 |
+|--------------------------|-----------------------------|
+| `/cluster/health`        | Cluster health metrics      |
+| `/cluster/usage/objects` | Object statistics           |
+| `/cluster/usage/buckets` | Object statistics by bucket |
+| `/cluster/erasure-set`   | Erasure set metrics         |
+|                          |                             |
+
+## Metrics Listing
+
+Each of the following sub-sections list metrics returned by each of the endpoints.
+
+The standard metrics groups for ProcessCollector and GoCollector are not shown below.
+
+### `/api/requests`
+
+| Name                                           | Type      | Help                                                    | Labels                           |
+|------------------------------------------------|-----------|---------------------------------------------------------|----------------------------------|
+| `minio_api_requests_rejected_auth_total`       | `counter` | Total number of requests rejected for auth failure      | `type,pool_index,server`         |
+| `minio_api_requests_rejected_header_total`     | `counter` | Total number of requests rejected for invalid header    | `type,pool_index,server`         |
+| `minio_api_requests_rejected_timestamp_total`  | `counter` | Total number of requests rejected for invalid timestamp | `type,pool_index,server`         |
+| `minio_api_requests_rejected_invalid_total`    | `counter` | Total number of invalid requests                        | `type,pool_index,server`         |
+| `minio_api_requests_waiting_total`             | `gauge`   | Total number of requests in the waiting queue           | `type,pool_index,server`         |
+| `minio_api_requests_incoming_total`            | `gauge`   | Total number of incoming requests                       | `type,pool_index,server`         |
+| `minio_api_requests_inflight_total`            | `gauge`   | Total number of requests currently in flight            | `name,type,pool_index,server`    |
+| `minio_api_requests_total`                     | `counter` | Total number of requests                                | `name,type,pool_index,server`    |
+| `minio_api_requests_errors_total`              | `counter` | Total number of requests with (4xx and 5xx) errors      | `name,type,pool_index,server`    |
+| `minio_api_requests_5xx_errors_total`          | `counter` | Total number of requests with 5xx errors                | `name,type,pool_index,server`    |
+| `minio_api_requests_4xx_errors_total`          | `counter` | Total number of requests with 4xx errors                | `name,type,pool_index,server`    |
+| `minio_api_requests_canceled_total`            | `counter` | Total number of requests canceled by the client         | `name,type,pool_index,server`    |
+| `minio_api_requests_ttfb_seconds_distribution` | `counter` | Distribution of time to first byte across API calls     | `name,type,le,pool_index,server` |
+| `minio_api_requests_traffic_sent_bytes`        | `counter` | Total number of bytes sent                              | `type,pool_index,server`         |
+| `minio_api_requests_traffic_received_bytes`    | `counter` | Total number of bytes received                          | `type,pool_index,server`         |
+
+### `/api/bucket`
+
+| Name                                         | Type      | Help                                                             | Labels                                  |
+|----------------------------------------------|-----------|------------------------------------------------------------------|-----------------------------------------|
+| `minio_api_bucket_traffic_received_bytes`    | `counter` | Total number of bytes sent for a bucket                          | `bucket,type,server,pool_index`         |
+| `minio_api_bucket_traffic_sent_bytes`        | `counter` | Total number of bytes received for a bucket                      | `bucket,type,server,pool_index`         |
+| `minio_api_bucket_inflight_total`            | `gauge`   | Total number of requests currently in flight for a bucket        | `bucket,name,type,server,pool_index`    |
+| `minio_api_bucket_total`                     | `counter` | Total number of requests for a bucket                            | `bucket,name,type,server,pool_index`    |
+| `minio_api_bucket_canceled_total`            | `counter` | Total number of requests canceled by the client for a bucket     | `bucket,name,type,server,pool_index`    |
+| `minio_api_bucket_4xx_errors_total`          | `counter` | Total number of requests with 4xx errors for a bucket            | `bucket,name,type,server,pool_index`    |
+| `minio_api_bucket_5xx_errors_total`          | `counter` | Total number of requests with 5xx errors for a bucket            | `bucket,name,type,server,pool_index`    |
+| `minio_api_bucket_ttfb_seconds_distribution` | `counter` | Distribution of time to first byte across API calls for a bucket | `bucket,name,le,type,server,pool_index` |
+
+### `/system/drive`
+
+| Name                                           | Type      | Help                                                                              | Labels                                              |
+|------------------------------------------------|-----------|-----------------------------------------------------------------------------------|-----------------------------------------------------|
+| `minio_system_drive_used_bytes`                | `gauge`   | Total storage used on a drive in bytes                                            | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_free_bytes`                | `gauge`   | Total storage free on a drive in bytes                                            | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_total_bytes`               | `gauge`   | Total storage available on a drive in bytes                                       | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_free_inodes`               | `gauge`   | Total free inodes on a drive                                                      | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_timeout_errors_total`      | `counter` | Total timeout errors on a drive                                                   | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_availability_errors_total` | `counter` | Total availability errors (I/O errors, permission denied and timeouts) on a drive | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_waiting_io`                | `gauge`   | Total waiting I/O operations on a drive                                           | `drive,set_index,drive_index,pool_index,server`     |
+| `minio_system_drive_api_latency_micros`        | `gauge`   | Average last minute latency in µs for drive API storage operations                | `drive,api,set_index,drive_index,pool_index,server` |
+| `minio_system_drive_offline_count`             | `gauge`   | Count of offline drives                                                           | `pool_index,server`                                 |
+| `minio_system_drive_online_count`              | `gauge`   | Count of online drives                                                            | `pool_index,server`                                 |
+| `minio_system_drive_count`                     | `gauge`   | Count of all drives                                                               | `pool_index,server`                                 |
+
+### `/system/network/internode`
+
+| Name                                                 | Type      | Help                                                     | Labels              |
+|------------------------------------------------------|-----------|----------------------------------------------------------|---------------------|
+| `minio_system_network_internode_errors_total`        | `counter` | Total number of failed internode calls                   | `server,pool_index` |
+| `minio_system_network_internode_dial_errors_total`   | `counter` | Total number of internode TCP dial timeouts and errors   | `server,pool_index` |
+| `minio_system_network_internode_dial_avg_time_nanos` | `gauge`   | Average dial time of internodes TCP calls in nanoseconds | `server,pool_index` |
+| `minio_system_network_internode_sent_bytes_total`    | `counter` | Total number of bytes sent to other peer nodes           | `server,pool_index` |
+| `minio_system_network_internode_recv_bytes_total`    | `counter` | Total number of bytes received from other peer nodes     | `server,pool_index` |
+
+### `/cluster/health`
+
+| Name                                               | Type    | Help                                           | Labels |
+|----------------------------------------------------|---------|------------------------------------------------|--------|
+| `minio_cluster_health_drives_offline_count`        | `gauge` | Count of offline drives in the cluster         |        |
+| `minio_cluster_health_drives_online_count`         | `gauge` | Count of online drives in the cluster          |        |
+| `minio_cluster_health_drives_count`                | `gauge` | Count of all drives in the cluster             |        |
+| `minio_cluster_health_nodes_offline_count`         | `gauge` | Count of offline nodes in the cluster          |        |
+| `minio_cluster_health_nodes_online_count`          | `gauge` | Count of online nodes in the cluster           |        |
+| `minio_cluster_health_capacity_raw_total_bytes`    | `gauge` | Total cluster raw storage capacity in bytes    |        |
+| `minio_cluster_health_capacity_raw_free_bytes`     | `gauge` | Total cluster raw storage free in bytes        |        |
+| `minio_cluster_health_capacity_usable_total_bytes` | `gauge` | Total cluster usable storage capacity in bytes |        |
+| `minio_cluster_health_capacity_usable_free_bytes`  | `gauge` | Total cluster usable storage free in bytes     |        |
+
+
+### `/cluster/usage/objects`
+
+| Name                                                     | Type    | Help                                                           | Labels  |
+|----------------------------------------------------------|---------|----------------------------------------------------------------|---------|
+| `minio_cluster_usage_objects_since_last_update_seconds`  | `gauge` | Time since last update of usage metrics in seconds             |         |
+| `minio_cluster_usage_objects_total_bytes`                | `gauge` | Total cluster usage in bytes                                   |         |
+| `minio_cluster_usage_objects_count`                      | `gauge` | Total cluster objects count                                    |         |
+| `minio_cluster_usage_objects_versions_count`             | `gauge` | Total cluster object versions (including delete markers) count |         |
+| `minio_cluster_usage_objects_delete_markers_count`       | `gauge` | Total cluster delete markers count                             |         |
+| `minio_cluster_usage_objects_buckets_count`              | `gauge` | Total cluster buckets count                                    |         |
+| `minio_cluster_usage_objects_size_distribution`          | `gauge` | Cluster object size distribution                               | `range` |
+| `minio_cluster_usage_objects_version_count_distribution` | `gauge` | Cluster object version count distribution                      | `range` |
+
+### `/cluster/usage/buckets` 
+
+| Name                                                            | Type    | Help                                                             | Labels         |
+|-----------------------------------------------------------------|---------|------------------------------------------------------------------|----------------|
+| `minio_cluster_usage_buckets_since_last_update_seconds`         | `gauge` | Time since last update of usage metrics in seconds               |                |
+| `minio_cluster_usage_buckets_total_bytes`                       | `gauge` | Total bucket size in bytes                                       | `bucket`       |
+| `minio_cluster_usage_buckets_objects_count`                     | `gauge` | Total objects count in bucket                                    | `bucket`       |
+| `minio_cluster_usage_buckets_versions_count`                    | `gauge` | Total object versions (including delete markers) count in bucket | `bucket`       |
+| `minio_cluster_usage_buckets_delete_markers_count`              | `gauge` | Total delete markers count in bucket                             | `bucket`       |
+| `minio_cluster_usage_buckets_quota_total_bytes`                 | `gauge` | Total bucket quota in bytes                                      | `bucket`       |
+| `minio_cluster_usage_buckets_object_size_distribution`          | `gauge` | Bucket object size distribution                                  | `range,bucket` |
+| `minio_cluster_usage_buckets_object_version_count_distribution` | `gauge` | Bucket object version count distribution                         | `range,bucket` |
+
+### `/cluster/erasure-set`
+
+| Name                                             | Type    | Help                                                          | Labels           |
+|--------------------------------------------------|---------|---------------------------------------------------------------|------------------|
+| `minio_cluster_erasure_set_overall_write_quorum` | `gauge` | Overall write quorum across pools and sets                    |                  |
+| `minio_cluster_erasure_set_overall_health`       | `gauge` | Overall health across pools and sets (1=healthy, 0=unhealthy) |                  |
+| `minio_cluster_erasure_set_read_quorum`          | `gauge` | Read quorum for the erasure set in a pool                     | `pool_id,set_id` |
+| `minio_cluster_erasure_set_write_quorum`         | `gauge` | Write quorum for the erasure set in a pool                    | `pool_id,set_id` |
+| `minio_cluster_erasure_set_online_drives_count`  | `gauge` | Count of online drives in the erasure set in a pool           | `pool_id,set_id` |
+| `minio_cluster_erasure_set_healing_drives_count` | `gauge` | Count of healing drives in the erasure set in a pool          | `pool_id,set_id` |
+| `minio_cluster_erasure_set_health`               | `gauge` | Health of the erasure set in a pool (1=healthy, 0=unhealthy)  | `pool_id,set_id` |