feat(inputs.internal): Add Go metric collection option (#13793)

plugins/inputs/internal/README.md

@@ -21,6 +21,10 @@ See the [CONFIGURATION.md][CONFIGURATION.md] for more details.
 [[inputs.internal]]
   ## If true, collect telegraf memory stats.
   # collect_memstats = true
+
+  ## If true, collect metrics from Go's runtime.metrics. For a full list see:
+  ##   https://pkg.go.dev/runtime/metrics
+  # collect_gostats = false
 ```
 
 ## Metrics

plugins/inputs/internal/internal.go

@@ -3,7 +3,9 @@ package internal
 
 import (
 	_ "embed"
+	"fmt"
 	"runtime"
+	"runtime/metrics"
 	"strings"
 
 	"github.com/influxdata/telegraf"
@@ -16,13 +18,8 @@ import (
 var sampleConfig string
 
 type Self struct {
-	CollectMemstats bool
-}
-
-func NewSelf() telegraf.Input {
-	return &Self{
-		CollectMemstats: true,
-	}
+	CollectMemstats bool `toml:"collect_memstats"`
+	CollectGostats  bool `toml:"collect_gostats"`
 }
 
 func (*Self) SampleConfig() string {
@@ -30,42 +27,116 @@ func (*Self) SampleConfig() string {
 }
 
 func (s *Self) Gather(acc telegraf.Accumulator) error {
-	if s.CollectMemstats {
-		m := &runtime.MemStats{}
-		runtime.ReadMemStats(m)
-		fields := map[string]interface{}{
-			"alloc_bytes":       m.Alloc,      // bytes allocated and not yet freed
-			"total_alloc_bytes": m.TotalAlloc, // bytes allocated (even if freed)
-			"sys_bytes":         m.Sys,        // bytes obtained from system (sum of XxxSys below)
-			"pointer_lookups":   m.Lookups,    // number of pointer lookups
-			"mallocs":           m.Mallocs,    // number of mallocs
-			"frees":             m.Frees,      // number of frees
-			// Main allocation heap statistics.
-			"heap_alloc_bytes":    m.HeapAlloc,    // bytes allocated and not yet freed (same as Alloc above)
-			"heap_sys_bytes":      m.HeapSys,      // bytes obtained from system
-			"heap_idle_bytes":     m.HeapIdle,     // bytes in idle spans
-			"heap_in_use_bytes":   m.HeapInuse,    // bytes in non-idle span
-			"heap_released_bytes": m.HeapReleased, // bytes released to the OS
-			"heap_objects":        m.HeapObjects,  // total number of allocated objects
-			"num_gc":              m.NumGC,
-		}
-		acc.AddFields("internal_memstats", fields, map[string]string{})
-	}
-
-	telegrafVersion := inter.Version
-	goVersion := strings.TrimPrefix(runtime.Version(), "go")
-
 	for _, m := range selfstat.Metrics() {
 		if m.Name() == "internal_agent" {
-			m.AddTag("go_version", goVersion)
+			m.AddTag("go_version", strings.TrimPrefix(runtime.Version(), "go"))
 		}
-		m.AddTag("version", telegrafVersion)
+		m.AddTag("version", inter.Version)
 		acc.AddFields(m.Name(), m.Fields(), m.Tags(), m.Time())
 	}
 
+	if s.CollectMemstats {
+		collectMemStat(acc)
+	}
+
+	if s.CollectGostats {
+		collectGoStat(acc)
+	}
+
 	return nil
 }
 
-func init() {
-	inputs.Add("internal", NewSelf)
+func collectMemStat(acc telegraf.Accumulator) {
+	m := &runtime.MemStats{}
+	runtime.ReadMemStats(m)
+	fields := map[string]any{
+		"alloc_bytes":       m.Alloc,      // bytes allocated and not yet freed
+		"total_alloc_bytes": m.TotalAlloc, // bytes allocated (even if freed)
+		"sys_bytes":         m.Sys,        // bytes obtained from system (sum of XxxSys below)
+		"pointer_lookups":   m.Lookups,    // number of pointer lookups
+		"mallocs":           m.Mallocs,    // number of mallocs
+		"frees":             m.Frees,      // number of frees
+
+		// Main allocation heap statistics.
+		"heap_alloc_bytes":    m.HeapAlloc,    // bytes allocated and not yet freed (same as Alloc above)
+		"heap_sys_bytes":      m.HeapSys,      // bytes obtained from system
+		"heap_idle_bytes":     m.HeapIdle,     // bytes in idle spans
+		"heap_in_use_bytes":   m.HeapInuse,    // bytes in non-idle span
+		"heap_released_bytes": m.HeapReleased, // bytes released to the OS
+		"heap_objects":        m.HeapObjects,  // total number of allocated objects
+		"num_gc":              m.NumGC,
+	}
+	acc.AddFields("internal_memstats", fields, map[string]string{})
+}
+
+func collectGoStat(acc telegraf.Accumulator) {
+	descs := metrics.All()
+	samples := make([]metrics.Sample, len(descs))
+	for i := range samples {
+		samples[i].Name = descs[i].Name
+	}
+	metrics.Read(samples)
+
+	fields := map[string]any{}
+	for _, sample := range samples {
+		name := sanitizeName(sample.Name)
+
+		switch sample.Value.Kind() {
+		case metrics.KindUint64:
+			fields[name] = sample.Value.Uint64()
+		case metrics.KindFloat64:
+			fields[name] = sample.Value.Float64()
+		case metrics.KindFloat64Histogram:
+			// The histogram may be quite large, so let's just pull out
+			// a crude estimate for the median for the sake of this example.
+			fields[name] = medianBucket(sample.Value.Float64Histogram())
+		default:
+			// This may happen as new metrics get added.
+			//
+			// The safest thing to do here is to simply log it somewhere
+			// as something to look into, but ignore it for now.
+			// In the worst case, you might temporarily miss out on a new metric.
+			fmt.Printf("%s: unexpected metric Kind: %v\n", name, sample.Value.Kind())
+		}
+	}
+
+	tags := map[string]string{
+		"go_version": strings.TrimPrefix(runtime.Version(), "go"),
+	}
+	acc.AddFields("internal_gostats", fields, tags)
+}
+
+// Converts /cpu/classes/gc/mark/assist:cpu-seconds to cpu_classes_gc_mark_assist_cpu_seconds
+func sanitizeName(name string) string {
+	name = strings.TrimPrefix(name, "/")
+	name = strings.ReplaceAll(name, "/", "_")
+	name = strings.ReplaceAll(name, ":", "_")
+	name = strings.ReplaceAll(name, "-", "_")
+	return name
+}
+
+func medianBucket(h *metrics.Float64Histogram) float64 {
+	total := uint64(0)
+	for _, count := range h.Counts {
+		total += count
+	}
+	thresh := total / 2
+	total = 0
+	for i, count := range h.Counts {
+		total += count
+		if total >= thresh {
+			return h.Buckets[i]
+		}
+	}
+
+	// default value in case something above did not work
+	return 0.0
+}
+
+func init() {
+	inputs.Add("internal", func() telegraf.Input {
+		return &Self{
+			CollectMemstats: true,
+		}
+	})
 }

plugins/inputs/internal/internal_test.go

@@ -1,6 +1,7 @@
 package internal
 
 import (
+	"fmt"
 	"testing"
 
 	"github.com/influxdata/telegraf/selfstat"
@@ -10,7 +11,9 @@ import (
 )
 
 func TestSelfPlugin(t *testing.T) {
-	s := NewSelf()
+	s := Self{
+		CollectMemstats: true,
+	}
 	acc := &testutil.Accumulator{}
 
 	require.NoError(t, s.Gather(acc))
@@ -64,3 +67,48 @@ func TestSelfPlugin(t *testing.T) {
 		},
 	)
 }
+
+func TestNoMemStat(t *testing.T) {
+	s := Self{
+		CollectMemstats: false,
+		CollectGostats:  false,
+	}
+	acc := &testutil.Accumulator{}
+
+	require.NoError(t, s.Gather(acc))
+	require.False(t, acc.HasMeasurement("internal_memstats"))
+	require.False(t, acc.HasMeasurement("internal_gostats"))
+}
+
+func TestGostats(t *testing.T) {
+	s := Self{
+		CollectMemstats: false,
+		CollectGostats:  true,
+	}
+	acc := &testutil.Accumulator{}
+
+	require.NoError(t, s.Gather(acc))
+	require.False(t, acc.HasMeasurement("internal_memstats"))
+	require.True(t, acc.HasMeasurement("internal_gostats"))
+
+	var metric *testutil.Metric
+	for _, m := range acc.Metrics {
+		if m.Measurement == "internal_gostats" {
+			metric = m
+			break
+		}
+	}
+
+	require.NotNil(t, metric)
+	require.Equal(t, metric.Measurement, "internal_gostats")
+	require.Equal(t, len(metric.Tags), 1)
+	require.Contains(t, metric.Tags, "go_version")
+
+	for name, value := range metric.Fields {
+		switch value.(type) {
+		case int64, uint64, float64:
+		default:
+			require.True(t, false, fmt.Sprintf("field %s is of non-numeric type %T\n", name, value))
+		}
+	}
+}

plugins/inputs/internal/sample.conf

@@ -2,3 +2,7 @@
 [[inputs.internal]]
   ## If true, collect telegraf memory stats.
   # collect_memstats = true
+
+  ## If true, collect metrics from Go's runtime.metrics. For a full list see:
+  ##   https://pkg.go.dev/runtime/metrics
+  # collect_gostats = false