Add basic scanner metrics (#13317)

Add number of objects/versions/folders scanned as well as ILM action outcomes.

cmd/data-scanner.go

@@ -29,6 +29,7 @@ import (
 	"path"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/bits-and-blooms/bloom/v3"
@@ -194,6 +195,22 @@ type folderScanner struct {
 	lastUpdate time.Time
 }
 
+type scannerStats struct {
+	// All fields must be accessed atomically and aligned.
+
+	accTotalObjects  uint64
+	accTotalVersions uint64
+	accFolders       uint64
+	bucketsStarted   uint64
+	bucketsFinished  uint64
+	ilmChecks        uint64
+
+	// actions records actions performed.
+	actions [lifecycle.ActionCount]uint64
+}
+
+var globalScannerStats scannerStats
+
 // Cache structure and compaction:
 //
 // A cache structure will be kept with a tree of usages.
@@ -241,6 +258,10 @@ func scanDataFolder(ctx context.Context, basePath string, cache dataUsageCache,
 
 	logPrefix := color.Green("data-usage: ")
 	logSuffix := color.Blue("- %v + %v", basePath, cache.Info.Name)
+	atomic.AddUint64(&globalScannerStats.bucketsStarted, 1)
+	defer func() {
+		atomic.AddUint64(&globalScannerStats.bucketsFinished, 1)
+	}()
 	if intDataUpdateTracker.debug {
 		defer func() {
 			console.Debugf(logPrefix+" Scanner time: %v %s\n", time.Since(t), logSuffix)
@@ -349,6 +370,7 @@ func (f *folderScanner) scanFolder(ctx context.Context, folder cachedFolder, int
 	thisHash := hashPath(folder.name)
 	// Store initial compaction state.
 	wasCompacted := into.Compacted
+	atomic.AddUint64(&globalScannerStats.accFolders, 1)
 
 	for {
 		select {
@@ -876,6 +898,7 @@ func (i *scannerItem) applyLifecycle(ctx context.Context, o ObjectLayer, oi Obje
 		return false, size
 	}
 
+	atomic.AddUint64(&globalScannerStats.ilmChecks, 1)
 	versionID := oi.VersionID
 	action := i.lifeCycle.ComputeAction(
 		lifecycle.ObjectOpts{
@@ -898,6 +921,8 @@ func (i *scannerItem) applyLifecycle(ctx context.Context, o ObjectLayer, oi Obje
 			console.Debugf(applyActionsLogPrefix+" lifecycle: %q Initial scan: %v\n", i.objectPath(), action)
 		}
 	}
+	atomic.AddUint64(&globalScannerStats.actions[action], 1)
+
 	switch action {
 	case lifecycle.DeleteAction, lifecycle.DeleteVersionAction, lifecycle.DeleteRestoredAction, lifecycle.DeleteRestoredVersionAction:
 		return applyLifecycleAction(action, oi), 0

cmd/fs-v1.go

@@ -393,6 +393,8 @@ func (fs *FSObjects) scanBucket(ctx context.Context, bucket string, cache dataUs
 		}
 
 		oi := fsMeta.ToObjectInfo(bucket, object, fi)
+		atomic.AddUint64(&globalScannerStats.accTotalVersions, 1)
+		atomic.AddUint64(&globalScannerStats.accTotalObjects, 1)
 		sz := item.applyActions(ctx, fs, oi, &sizeSummary{})
 		if sz >= 0 {
 			return sizeSummary{totalSize: sz, versions: 1}, nil

cmd/metrics-v2.go

@@ -24,8 +24,10 @@ import (
 	"runtime"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
+	"github.com/minio/minio/internal/bucket/lifecycle"
 	"github.com/minio/minio/internal/logger"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
@@ -69,6 +71,7 @@ const (
 	sysCallSubsystem          MetricSubsystem = "syscall"
 	usageSubsystem            MetricSubsystem = "usage"
 	ilmSubsystem              MetricSubsystem = "ilm"
+	scannerSubsystem          MetricSubsystem = "scanner"
 )
 
 // MetricName are the individual names for the metric.
@@ -255,6 +258,7 @@ func GetGeneratorsForPeer() []MetricsGenerator {
 		getNetworkMetrics,
 		getS3TTFBMetric,
 		getILMNodeMetrics,
+		getScannerNodeMetrics,
 	}
 	return g
 }
@@ -1066,6 +1070,95 @@ func getILMNodeMetrics() MetricsGroup {
 	}
 }
 
+func getScannerNodeMetrics() MetricsGroup {
+	return MetricsGroup{
+		id:         "ScannerNodeMetrics",
+		cachedRead: cachedRead,
+		read: func(_ context.Context) []Metric {
+			metrics := []Metric{
+				{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: scannerSubsystem,
+						Name:      "objects_scanned",
+						Help:      "Total number of unique objects scanned since server start.",
+						Type:      counterMetric,
+					},
+					Value: float64(atomic.LoadUint64(&globalScannerStats.accTotalObjects)),
+				},
+				{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: scannerSubsystem,
+						Name:      "versions_scanned",
+						Help:      "Total number of object versions scanned since server start.",
+						Type:      counterMetric,
+					},
+					Value: float64(atomic.LoadUint64(&globalScannerStats.accTotalVersions)),
+				},
+				{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: scannerSubsystem,
+						Name:      "directories_scanned",
+						Help:      "Total number of directories scanned since server start.",
+						Type:      counterMetric,
+					},
+					Value: float64(atomic.LoadUint64(&globalScannerStats.accFolders)),
+				},
+				{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: scannerSubsystem,
+						Name:      "bucket_scans_started",
+						Help:      "Total number of bucket scans started since server start",
+						Type:      counterMetric,
+					},
+					Value: float64(atomic.LoadUint64(&globalScannerStats.bucketsStarted)),
+				},
+				{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: scannerSubsystem,
+						Name:      "bucket_scans_finished",
+						Help:      "Total number of bucket scans finished since server start",
+						Type:      counterMetric,
+					},
+					Value: float64(atomic.LoadUint64(&globalScannerStats.bucketsFinished)),
+				},
+				{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: ilmSubsystem,
+						Name:      "versions_scanned",
+						Help:      "Total number of object versions checked for ilm actions since server start",
+						Type:      counterMetric,
+					},
+					Value: float64(atomic.LoadUint64(&globalScannerStats.ilmChecks)),
+				},
+			}
+			for i := range globalScannerStats.actions {
+				action := lifecycle.Action(i)
+				v := atomic.LoadUint64(&globalScannerStats.actions[action])
+				if v == 0 {
+					continue
+				}
+				metrics = append(metrics, Metric{
+					Description: MetricDescription{
+						Namespace: nodeMetricNamespace,
+						Subsystem: ilmSubsystem,
+						Name:      MetricName("action_count_" + toSnake(action.String())),
+						Help:      "Total action outcome of lifecycle checks since server start",
+						Type:      counterMetric,
+					},
+					Value: float64(v),
+				})
+			}
+			return metrics
+		},
+	}
+}
+
 func getMinioVersionMetrics() MetricsGroup {
 	return MetricsGroup{
 		id:         "MinioVersionMetrics",
@@ -1753,3 +1846,26 @@ func metricsNodeHandler() http.Handler {
 			}),
 	)
 }
+
+func toSnake(camel string) (snake string) {
+	var b strings.Builder
+	l := len(camel)
+	for i, v := range camel {
+		// A is 65, a is 97
+		if v >= 'a' {
+			b.WriteRune(v)
+			continue
+		}
+		// v is capital letter here
+		// disregard first letter
+		// add underscore if last letter is capital letter
+		// add underscore when previous letter is lowercase
+		// add underscore when next letter is lowercase
+		if (i != 0 || i == l-1) && ((i > 0 && rune(camel[i-1]) >= 'a') ||
+			(i < l-1 && rune(camel[i+1]) >= 'a')) {
+			b.WriteRune('_')
+		}
+		b.WriteRune(v + 'a' - 'A')
+	}
+	return b.String()
+}

cmd/xl-storage.go

@@ -33,6 +33,7 @@ import (
 	"runtime"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"syscall"
 	"time"
 
@@ -462,7 +463,9 @@ func (s *xlStorage) NSScanner(ctx context.Context, cache dataUsageCache, updates
 			return sizeSummary{}, errSkipFile
 		}
 		sizeS := sizeSummary{}
+		atomic.AddUint64(&globalScannerStats.accTotalObjects, 1)
 		for _, version := range fivs.Versions {
+			atomic.AddUint64(&globalScannerStats.accTotalVersions, 1)
 			oi := version.ToObjectInfo(item.bucket, item.objectPath())
 			sz := item.applyActions(ctx, objAPI, oi, &sizeS)
 			if !oi.DeleteMarker && sz == oi.Size {

internal/bucket/lifecycle/action_string.go

@@ -15,11 +15,12 @@ func _() {
 	_ = x[TransitionVersionAction-4]
 	_ = x[DeleteRestoredAction-5]
 	_ = x[DeleteRestoredVersionAction-6]
+	_ = x[ActionCount-7]
 }
 
-const _Action_name = "NoneActionDeleteActionDeleteVersionActionTransitionActionTransitionVersionActionDeleteRestoredActionDeleteRestoredVersionAction"
+const _Action_name = "NoneActionDeleteActionDeleteVersionActionTransitionActionTransitionVersionActionDeleteRestoredActionDeleteRestoredVersionActionActionCount"
 
-var _Action_index = [...]uint8{0, 10, 22, 41, 57, 80, 100, 127}
+var _Action_index = [...]uint8{0, 10, 22, 41, 57, 80, 100, 127, 138}
 
 func (i Action) String() string {
 	if i < 0 || i >= Action(len(_Action_index)-1) {

internal/bucket/lifecycle/lifecycle.go

@@ -63,6 +63,9 @@ const (
 	DeleteRestoredAction
 	// DeleteRestoredVersionAction deletes a particular version that was temporarily restored
 	DeleteRestoredVersionAction
+
+	// ActionCount must be the last action and shouldn't be used as a regular action.
+	ActionCount
 )
 
 // Lifecycle - Configuration for bucket lifecycle.