metrics/histogram.go
2019-11-23 23:58:18 +02:00

262 lines
6.2 KiB
Go

package metrics
import (
"fmt"
"io"
"math"
"sync"
"sync/atomic"
"time"
)
// Histogram is a histogram that covers values with the following buckets:
//
// 0
// (0...1e-9]
// (1e-9...2e-9]
// (2e-9...3e-9]
// ...
// (9e-9...1e-8]
// (1e-8...2e-8]
// ...
// (1e11...2e11]
// (2e11...3e11]
// ...
// (9e11...1e12]
// (1e12...Inf]
//
// Each bucket contains a counter for values in the given range.
// Each non-zero bucket is exposed with the following name:
//
// <metric_name>_bucket{<optional_tags>,vmrange="<start>...<end>"} <counter>
//
// Where:
//
// - <metric_name> is the metric name passed to NewHistogram
// - <optional_tags> is optional tags for the <metric_name>, which are passed to NewHistogram
// - <start> and <end> - start and end values for the given bucket
// - <counter> - the number of hits to the given bucket during Update* calls.
//
// Histogram buckets can be converted to Prometheus-like buckets with `le` labels
// with `prometheus_buckets(<metric_name>_bucket)` function in VictoriaMetrics:
//
// prometheus_buckets(request_duration_bucket)
//
// Histogram cannot be used for negative values.
//
// Time series produced by the Histogram have better compression ratio comparing to
// Prometheus histogram buckets with `le` labels, since they don't include counters
// for all the previous buckets.
type Histogram struct {
buckets [bucketsCount]uint64
sumMu sync.Mutex
sum float64
count uint64
}
// NewHistogram creates and returns new histogram with the given name.
//
// name must be valid Prometheus-compatible metric with possible labels.
// For instance,
//
// * foo
// * foo{bar="baz"}
// * foo{bar="baz",aaa="b"}
//
// The returned histogram is safe to use from concurrent goroutines.
func NewHistogram(name string) *Histogram {
return defaultSet.NewHistogram(name)
}
// GetOrCreateHistogram returns registered histogram with the given name
// or creates new histogram if the registry doesn't contain histogram with
// the given name.
//
// name must be valid Prometheus-compatible metric with possible labels.
// For instance,
//
// * foo
// * foo{bar="baz"}
// * foo{bar="baz",aaa="b"}
//
// The returned histogram is safe to use from concurrent goroutines.
//
// Performance tip: prefer NewHistogram instead of GetOrCreateHistogram.
func GetOrCreateHistogram(name string) *Histogram {
return defaultSet.GetOrCreateHistogram(name)
}
// Update updates h with v.
//
// v cannot be negative.
func (h *Histogram) Update(v float64) {
if math.IsNaN(v) || v < 0 {
// Skip NaNs and negative values.
return
}
idx := getBucketIdx(v)
if idx >= uint(len(h.buckets)) {
panic(fmt.Errorf("BUG: idx cannot exceed %d; got %d", len(h.buckets), idx))
}
atomic.AddUint64(&h.buckets[idx], 1)
atomic.AddUint64(&h.count, 1)
h.sumMu.Lock()
h.sum += v
h.sumMu.Unlock()
}
// UpdateDuration updates request duration based on the given startTime.
func (h *Histogram) UpdateDuration(startTime time.Time) {
d := time.Since(startTime).Seconds()
h.Update(d)
}
// VisitNonZeroBuckets calls f for all buckets with non-zero counters.
func (h *Histogram) VisitNonZeroBuckets(f func(vmrange string, count uint64)) {
for i, v := range h.buckets[:] {
if v == 0 {
continue
}
vmrange := getRangeForBucketIdx(uint(i))
f(vmrange, v)
}
}
func getRangeForBucketIdx(idx uint) string {
bucketRangesOnce.Do(initBucketRanges)
return bucketRanges[idx]
}
func initBucketRanges() {
start := "0"
for i := 0; i < bucketsCount; i++ {
end := getRangeEndFromBucketIdx(uint(i))
bucketRanges[i] = start + "..." + end
start = end
}
}
var (
bucketRanges [bucketsCount]string
bucketRangesOnce sync.Once
)
func getTagForBucketIdx(idx uint) string {
bucketTagsOnce.Do(initBucketTags)
return bucketTags[idx]
}
func initBucketTags() {
for i := 0; i < bucketsCount; i++ {
vmrange := getRangeForBucketIdx(uint(i))
bucketTags[i] = fmt.Sprintf(`vmrange=%q`, vmrange)
}
}
var (
bucketTags [bucketsCount]string
bucketTagsOnce sync.Once
)
func (h *Histogram) marshalTo(prefix string, w io.Writer) {
count := atomic.LoadUint64(&h.count)
if count == 0 {
return
}
for i := range h.buckets[:] {
h.marshalBucket(prefix, w, uint(i))
}
// Marshal `_sum` and `_count` metrics.
name, filters := splitMetricName(prefix)
h.sumMu.Lock()
sum := h.sum
h.sumMu.Unlock()
if float64(int64(sum)) == sum {
fmt.Fprintf(w, "%s_sum%s %d\n", name, filters, int64(sum))
} else {
fmt.Fprintf(w, "%s_sum%s %g\n", name, filters, sum)
}
fmt.Fprintf(w, "%s_count%s %d\n", name, filters, count)
}
func (h *Histogram) marshalBucket(prefix string, w io.Writer, idx uint) {
v := h.buckets[idx]
if v == 0 {
return
}
tag := getTagForBucketIdx(idx)
prefix = addTag(prefix, tag)
name, filters := splitMetricName(prefix)
fmt.Fprintf(w, "%s_bucket%s %d\n", name, filters, v)
}
func getBucketIdx(v float64) uint {
if v < 0 {
panic(fmt.Errorf("BUG: v cannot be negative; got %v", v))
}
if v == 0 {
// Fast path for zero.
return 0
}
if math.IsInf(v, 1) {
return bucketsCount - 1
}
e10 := int(math.Floor(math.Log10(v)))
if e10 < e10Min {
return 1
}
if e10 > e10Max {
if e10 == e10Max+1 && math.Pow10(e10) == v {
// Adjust m to be on par with Prometheus 'le' buckets (aka 'less or equal')
return bucketsCount - 2
}
return bucketsCount - 1
}
mf := v / math.Pow10(e10)
m := uint(mf)
// Handle possible rounding errors
if m < 1 {
m = 1
} else if m > 9 {
m = 9
}
if float64(m) == mf {
// Adjust m to be on par with Prometheus 'le' buckets (aka 'less or equal')
m--
}
return 1 + m + uint(e10-e10Min)*9
}
func getRangeEndFromBucketIdx(idx uint) string {
if idx == 0 {
return "0"
}
if idx == 1 {
return fmt.Sprintf("1e%d", e10Min)
}
if idx >= bucketsCount-1 {
return "+Inf"
}
idx -= 2
e10 := e10Min + int(idx/9)
m := 2 + (idx % 9)
if m == 10 {
e10++
m = 1
}
if e10 == 0 {
return fmt.Sprintf("%d", m)
}
return fmt.Sprintf("%de%d", m, e10)
}
// Each range (10^n..10^(n+1)] for e10Min<=n<=e10Max is split into 9 equal sub-ranges, plus 3 additional buckets:
// - a bucket for zeros
// - a bucket for the range (0..10^e10Min]
// - a bucket for the range (10^(e10Max+1)..Inf]
const bucketsCount = 3 + 9*(1+e10Max-e10Min)
const e10Min = -9
const e10Max = 11