Export tier metrics (#18678)

minio_node_tier_ttlb_seconds - Distribution of time to last byte for streaming objects from warm tier
minio_node_tier_requests_success - Number of requests to download object from warm tier that were successful
minio_node_tier_requests_failure - Number of requests to download object from warm tier that failed

cmd/bucket-lifecycle.go

@@ -507,9 +507,13 @@ func auditTierActions(ctx context.Context, tier string, bytes int64) func(err er
 		}
 
 		if err == nil {
-			op.TimeToResponseNS = time.Since(startTime).Nanoseconds()
+			since := time.Since(startTime)
+			op.TimeToResponseNS = since.Nanoseconds()
+			globalTierMetrics.Observe(tier, since)
+			globalTierMetrics.logSuccess(tier)
 		} else {
 			op.Error = err.Error()
+			globalTierMetrics.logFailure(tier)
 		}
 
 		logger.GetReqInfo(ctx).AppendTags("tierStats", op)

cmd/metrics-v2.go

@@ -90,6 +90,7 @@ func init() {
 		getNetworkMetrics(),
 		getMinioVersionMetrics(),
 		getS3TTFBMetric(),
+		getTierMetrics(),
 		getNotificationMetrics(),
 		getDistLockMetrics(),
 		getIAMNodeMetrics(),
@@ -155,6 +156,7 @@ const (
 	usageSubsystem            MetricSubsystem = "usage"
 	quotaSubsystem            MetricSubsystem = "quota"
 	ilmSubsystem              MetricSubsystem = "ilm"
+	tierSubsystem             MetricSubsystem = "tier"
 	scannerSubsystem          MetricSubsystem = "scanner"
 	iamSubsystem              MetricSubsystem = "iam"
 	kmsSubsystem              MetricSubsystem = "kms"
@@ -246,6 +248,7 @@ const (
 	sizeDistribution    = "size_distribution"
 	versionDistribution = "version_distribution"
 	ttfbDistribution    = "seconds_distribution"
+	ttlbDistribution    = "ttlb_seconds_distribution"
 
 	lastActivityTime = "last_activity_nano_seconds"
 	startTime        = "starttime_seconds"
@@ -262,6 +265,9 @@ const (
 	transitionedObjects  MetricName = "transitioned_objects"
 	transitionedVersions MetricName = "transitioned_versions"
 
+	tierRequestsSuccess MetricName = "requests_success"
+	tierRequestsFailure MetricName = "requests_failure"
+
 	kmsOnline          = "online"
 	kmsRequestsSuccess = "request_success"
 	kmsRequestsError   = "request_error"
@@ -1658,6 +1664,16 @@ func getS3TTFBMetric() *MetricsGroup {
 	return mg
 }
 
+func getTierMetrics() *MetricsGroup {
+	mg := &MetricsGroup{
+		cacheInterval: 10 * time.Second,
+	}
+	mg.RegisterRead(func(ctx context.Context) []Metric {
+		return globalTierMetrics.Report()
+	})
+	return mg
+}
+
 func getTransitionPendingTasksMD() MetricDescription {
 	return MetricDescription{
 		Namespace: nodeMetricNamespace,

cmd/tier.go

@@ -1,4 +1,4 @@
-// Copyright (c) 2015-2021 MinIO, Inc.
+// Copyright (c) 2015-2023 MinIO, Inc.
 //
 // This file is part of MinIO Object Storage stack
 //
@@ -27,11 +27,13 @@ import (
 	"path"
 	"strings"
 	"sync"
+	"time"
 
 	"github.com/minio/madmin-go/v3"
 	"github.com/minio/minio/internal/crypto"
 	"github.com/minio/minio/internal/hash"
 	"github.com/minio/minio/internal/kms"
+	"github.com/prometheus/client_golang/prometheus"
 )
 
 //go:generate msgp -file $GOFILE
@@ -80,6 +82,96 @@ type TierConfigMgr struct {
 	Tiers map[string]madmin.TierConfig `json:"tiers"`
 }
 
+type tierMetrics struct {
+	sync.RWMutex  // protects requestsCount only
+	requestsCount map[string]struct {
+		success int64
+		failure int64
+	}
+	histogram *prometheus.HistogramVec
+}
+
+var globalTierMetrics = tierMetrics{
+	requestsCount: make(map[string]struct {
+		success int64
+		failure int64
+	}),
+	histogram: prometheus.NewHistogramVec(prometheus.HistogramOpts{
+		Name:    "tier_ttlb_seconds",
+		Help:    "Time taken by requests served by warm tier",
+		Buckets: []float64{0.01, 0.1, 1, 2, 5, 10, 60, 5 * 60, 15 * 60, 30 * 60},
+	}, []string{"tier"}),
+}
+
+func (t *tierMetrics) Observe(tier string, dur time.Duration) {
+	t.histogram.With(prometheus.Labels{"tier": tier}).Observe(dur.Seconds())
+}
+
+func (t *tierMetrics) logSuccess(tier string) {
+	t.Lock()
+	defer t.Unlock()
+
+	stat := t.requestsCount[tier]
+	stat.success++
+	t.requestsCount[tier] = stat
+}
+
+func (t *tierMetrics) logFailure(tier string) {
+	t.Lock()
+	defer t.Unlock()
+
+	stat := t.requestsCount[tier]
+	stat.failure++
+	t.requestsCount[tier] = stat
+}
+
+var (
+	// {minio_node}_{tier}_{ttlb_seconds_distribution}
+	tierTTLBMD = MetricDescription{
+		Namespace: nodeMetricNamespace,
+		Subsystem: tierSubsystem,
+		Name:      ttlbDistribution,
+		Help:      "Distribution of time to last byte for objects downloaded from warm tier",
+		Type:      gaugeMetric,
+	}
+
+	// {minio_node}_{tier}_{requests_success}
+	tierRequestsSuccessMD = MetricDescription{
+		Namespace: nodeMetricNamespace,
+		Subsystem: tierSubsystem,
+		Name:      tierRequestsSuccess,
+		Help:      "Number of requests to download object from warm tier that were successful",
+		Type:      counterMetric,
+	}
+	// {minio_node}_{tier}_{requests_failure}
+	tierRequestsFailureMD = MetricDescription{
+		Namespace: nodeMetricNamespace,
+		Subsystem: tierSubsystem,
+		Name:      tierRequestsFailure,
+		Help:      "Number of requests to download object from warm tier that failed",
+		Type:      counterMetric,
+	}
+)
+
+func (t *tierMetrics) Report() []Metric {
+	metrics := getHistogramMetrics(t.histogram, tierTTLBMD)
+	t.RLock()
+	defer t.RUnlock()
+	for tier, stat := range t.requestsCount {
+		metrics = append(metrics, Metric{
+			Description:    tierRequestsSuccessMD,
+			Value:          float64(stat.success),
+			VariableLabels: map[string]string{"tier": tier},
+		})
+		metrics = append(metrics, Metric{
+			Description:    tierRequestsFailureMD,
+			Value:          float64(stat.failure),
+			VariableLabels: map[string]string{"tier": tier},
+		})
+	}
+	return metrics
+}
+
 // IsTierValid returns true if there exists a remote tier by name tierName,
 // otherwise returns false.
 func (config *TierConfigMgr) IsTierValid(tierName string) bool {

cmd/tier_test.go

@@ -0,0 +1,52 @@
+// Copyright (c) 2015-2023 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 (
+	"testing"
+	"time"
+)
+
+func TestTierMetrics(t *testing.T) {
+	tier := "WARM-1"
+	globalTierMetrics.Observe(tier, 200*time.Millisecond)
+	expSuccess := 10
+	expFailure := 5
+	for i := 0; i < expSuccess; i++ {
+		globalTierMetrics.logSuccess(tier)
+	}
+	for i := 0; i < expFailure; i++ {
+		globalTierMetrics.logFailure(tier)
+	}
+	metrics := globalTierMetrics.Report()
+	var succ, fail float64
+	for _, metric := range metrics {
+		switch metric.Description.Name {
+		case tierRequestsSuccess:
+			succ += metric.Value
+		case tierRequestsFailure:
+			fail += metric.Value
+		}
+	}
+	if int(succ) != expSuccess {
+		t.Fatalf("Expected %d successes but got %f", expSuccess, succ)
+	}
+	if int(fail) != expFailure {
+		t.Fatalf("Expected %d failures but got %f", expFailure, fail)
+	}
+}

docs/metrics/prometheus/list.md

@@ -200,6 +200,14 @@ For deployments with [bucket](https://min.io/docs/minio/linux/administration/buc
 | `minio_node_ilm_transition_missed_immediate_tasks` | Number of missed immediate ILM transition tasks.                            |
 | `minio_node_ilm_versions_scanned`                  | Total number of object versions checked for ilm actions since server start. |
 
+## Tier Metrics
+
+| Name                                               | Description                                                                 |
+|:---------------------------------------------------|:----------------------------------------------------------------------------|
+| `minio_node_tier_tier_ttlb_seconds_distribution`   | Distribution of time to last byte for objects downloaded from warm tier     |
+| `minio_node_tier_requests_success`                 | Number of requests to download object from warm tier that were successful   | 
+| `minio_node_tier_requests_failure`                 | Number of requests to download object from warm tier that were failure      | 
+
 ## System Metrics
 
 | Name                                       | Description                                                                                                     |