VictoriaMetrics/lib/storage/metric_name.go
Aliaksandr Valialkin f4989edd96
lib/bytesutil: split Resize() into ResizeNoCopy() and ResizeWithCopy() functions
Previously bytesutil.Resize() was copying the original byte slice contents to a newly allocated slice.
This wasted CPU cycles and memory bandwidth in some places, where the original slice contents wasn't needed
after slize resizing. Switch such places to bytesutil.ResizeNoCopy().

Rename the original bytesutil.Resize() function to bytesutil.ResizeWithCopy() for the sake of improved readability.

Additionally, allocate new slice with `make()` instead of `append()`. This guarantees that the capacity of the allocated slice
exactly matches the requested size. The `append()` could return a slice with bigger capacity as an optimization for further `append()` calls.
This could result in excess memory usage when the returned byte slice was cached (for instance, in lib/blockcache).

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2007
2022-01-25 15:24:44 +02:00

763 lines
19 KiB
Go

package storage
import (
"bytes"
"fmt"
"sort"
"strconv"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
)
const (
escapeChar = 0
tagSeparatorChar = 1
kvSeparatorChar = 2
)
// Tag represents a (key, value) tag for metric.
type Tag struct {
Key []byte
Value []byte
}
// Reset resets the tag.
func (tag *Tag) Reset() {
tag.Key = tag.Key[:0]
tag.Value = tag.Value[:0]
}
// Equal returns true if tag equals t
func (tag *Tag) Equal(t *Tag) bool {
return string(tag.Key) == string(t.Key) && string(tag.Value) == string(t.Value)
}
// Marshal appends marshaled tag to dst and returns the result.
func (tag *Tag) Marshal(dst []byte) []byte {
dst = marshalTagValue(dst, tag.Key)
dst = marshalTagValue(dst, tag.Value)
return dst
}
// Unmarshal unmarshals tag from src and returns the remaining data from src.
func (tag *Tag) Unmarshal(src []byte) ([]byte, error) {
var err error
src, tag.Key, err = unmarshalTagValue(tag.Key[:0], src)
if err != nil {
return src, fmt.Errorf("cannot unmarshal key: %w", err)
}
src, tag.Value, err = unmarshalTagValue(tag.Value[:0], src)
if err != nil {
return src, fmt.Errorf("cannot unmarshal value: %w", err)
}
return src, nil
}
func (tag *Tag) copyFrom(src *Tag) {
tag.Key = append(tag.Key[:0], src.Key...)
tag.Value = append(tag.Value[:0], src.Value...)
}
func marshalTagValueNoTrailingTagSeparator(dst, src []byte) []byte {
dst = marshalTagValue(dst, src)
// Remove trailing tagSeparatorChar
return dst[:len(dst)-1]
}
func marshalTagValue(dst, src []byte) []byte {
n1 := bytes.IndexByte(src, escapeChar)
n2 := bytes.IndexByte(src, tagSeparatorChar)
n3 := bytes.IndexByte(src, kvSeparatorChar)
if n1 < 0 && n2 < 0 && n3 < 0 {
// Fast path.
dst = append(dst, src...)
dst = append(dst, tagSeparatorChar)
return dst
}
// Slow path.
for _, ch := range src {
switch ch {
case escapeChar:
dst = append(dst, escapeChar, '0')
case tagSeparatorChar:
dst = append(dst, escapeChar, '1')
case kvSeparatorChar:
dst = append(dst, escapeChar, '2')
default:
dst = append(dst, ch)
}
}
dst = append(dst, tagSeparatorChar)
return dst
}
func unmarshalTagValue(dst, src []byte) ([]byte, []byte, error) {
n := bytes.IndexByte(src, tagSeparatorChar)
if n < 0 {
return src, dst, fmt.Errorf("cannot find the end of tag value")
}
b := src[:n]
src = src[n+1:]
for {
n := bytes.IndexByte(b, escapeChar)
if n < 0 {
dst = append(dst, b...)
return src, dst, nil
}
dst = append(dst, b[:n]...)
b = b[n+1:]
if len(b) == 0 {
return src, dst, fmt.Errorf("missing escaped char")
}
switch b[0] {
case '0':
dst = append(dst, escapeChar)
case '1':
dst = append(dst, tagSeparatorChar)
case '2':
dst = append(dst, kvSeparatorChar)
default:
return src, dst, fmt.Errorf("unsupported escaped char: %c", b[0])
}
b = b[1:]
}
}
// MetricName reperesents a metric name.
type MetricName struct {
MetricGroup []byte
// Tags are optional. They must be sorted by tag Key for canonical view.
// Use sortTags method.
Tags []Tag
}
// GetMetricName returns a MetricName from pool.
func GetMetricName() *MetricName {
v := mnPool.Get()
if v == nil {
return &MetricName{}
}
return v.(*MetricName)
}
// PutMetricName returns mn to the pool.
func PutMetricName(mn *MetricName) {
mn.Reset()
mnPool.Put(mn)
}
var mnPool sync.Pool
// Reset resets the mn.
func (mn *MetricName) Reset() {
mn.MetricGroup = mn.MetricGroup[:0]
mn.Tags = mn.Tags[:0]
}
// CopyFrom copies src to mn.
func (mn *MetricName) CopyFrom(src *MetricName) {
if cap(mn.MetricGroup) > 0 {
mn.MetricGroup = append(mn.MetricGroup[:0], src.MetricGroup...)
mn.Tags = copyTags(mn.Tags[:0], src.Tags)
return
}
// Pre-allocate a single byte slice for MetricGroup + all the tags.
// This reduces the number of memory allocations for zero mn.
size := len(src.MetricGroup)
for i := range src.Tags {
tag := &src.Tags[i]
size += len(tag.Key)
size += len(tag.Value)
}
b := make([]byte, 0, size)
b = append(b, src.MetricGroup...)
mn.MetricGroup = b[:len(b):len(b)]
mn.Tags = make([]Tag, len(src.Tags))
for i := range src.Tags {
st := &src.Tags[i]
dt := &mn.Tags[i]
b = append(b, st.Key...)
dt.Key = b[len(b)-len(st.Key) : len(b) : len(b)]
b = append(b, st.Value...)
dt.Value = b[len(b)-len(st.Value) : len(b) : len(b)]
}
}
// AddTag adds new tag to mn with the given key and value.
func (mn *MetricName) AddTag(key, value string) {
if key == string(metricGroupTagKey) {
mn.MetricGroup = append(mn.MetricGroup, value...)
return
}
tag := mn.addNextTag()
tag.Key = append(tag.Key[:0], key...)
tag.Value = append(tag.Value[:0], value...)
}
// AddTagBytes adds new tag to mn with the given key and value.
func (mn *MetricName) AddTagBytes(key, value []byte) {
if string(key) == string(metricGroupTagKey) {
mn.MetricGroup = append(mn.MetricGroup, value...)
return
}
tag := mn.addNextTag()
tag.Key = append(tag.Key[:0], key...)
tag.Value = append(tag.Value[:0], value...)
}
func (mn *MetricName) addNextTag() *Tag {
if len(mn.Tags) < cap(mn.Tags) {
mn.Tags = mn.Tags[:len(mn.Tags)+1]
} else {
mn.Tags = append(mn.Tags, Tag{})
}
return &mn.Tags[len(mn.Tags)-1]
}
// ResetMetricGroup resets mn.MetricGroup
func (mn *MetricName) ResetMetricGroup() {
mn.MetricGroup = mn.MetricGroup[:0]
}
var metricGroupTagKey = []byte("__name__")
// RemoveTagsOn removes all the tags not included to onTags.
func (mn *MetricName) RemoveTagsOn(onTags []string) {
if !hasTag(onTags, metricGroupTagKey) {
mn.ResetMetricGroup()
}
tags := mn.Tags
mn.Tags = mn.Tags[:0]
if len(onTags) == 0 {
return
}
for i := range tags {
tag := &tags[i]
if hasTag(onTags, tag.Key) {
mn.AddTagBytes(tag.Key, tag.Value)
}
}
}
// RemoveTag removes a tag with the given tagKey
func (mn *MetricName) RemoveTag(tagKey string) {
if tagKey == "__name__" {
mn.ResetMetricGroup()
return
}
tags := mn.Tags
mn.Tags = mn.Tags[:0]
for i := range tags {
tag := &tags[i]
if string(tag.Key) != tagKey {
mn.AddTagBytes(tag.Key, tag.Value)
}
}
}
// RemoveTagsIgnoring removes all the tags included in ignoringTags.
func (mn *MetricName) RemoveTagsIgnoring(ignoringTags []string) {
if len(ignoringTags) == 0 {
return
}
if hasTag(ignoringTags, metricGroupTagKey) {
mn.ResetMetricGroup()
}
tags := mn.Tags
mn.Tags = mn.Tags[:0]
for i := range tags {
tag := &tags[i]
if !hasTag(ignoringTags, tag.Key) {
mn.AddTagBytes(tag.Key, tag.Value)
}
}
}
// GetTagValue returns tag value for the given tagKey.
func (mn *MetricName) GetTagValue(tagKey string) []byte {
if tagKey == "__name__" {
return mn.MetricGroup
}
tags := mn.Tags
for i := range tags {
tag := &tags[i]
if string(tag.Key) == tagKey {
return tag.Value
}
}
return nil
}
// SetTags sets tags from src with keys matching addTags.
func (mn *MetricName) SetTags(addTags []string, src *MetricName) {
for _, tagName := range addTags {
if tagName == string(metricGroupTagKey) {
mn.MetricGroup = append(mn.MetricGroup[:0], src.MetricGroup...)
continue
}
var srcTag *Tag
for i := range src.Tags {
t := &src.Tags[i]
if string(t.Key) == tagName {
srcTag = t
break
}
}
if srcTag == nil {
mn.RemoveTag(tagName)
continue
}
found := false
for i := range mn.Tags {
t := &mn.Tags[i]
if string(t.Key) == tagName {
t.Value = append(t.Value[:0], srcTag.Value...)
found = true
break
}
}
if !found {
mn.AddTagBytes(srcTag.Key, srcTag.Value)
}
}
}
func hasTag(tags []string, key []byte) bool {
for _, t := range tags {
if t == string(key) {
return true
}
}
return false
}
// String returns user-readable representation of the metric name.
func (mn *MetricName) String() string {
var mnCopy MetricName
mnCopy.CopyFrom(mn)
mnCopy.sortTags()
var tags []string
for i := range mnCopy.Tags {
t := &mnCopy.Tags[i]
tags = append(tags, fmt.Sprintf("%s=%q", t.Key, t.Value))
}
tagsStr := strings.Join(tags, ",")
return fmt.Sprintf("%s{%s}", mnCopy.MetricGroup, tagsStr)
}
// Marshal appends marshaled mn to dst and returns the result.
//
// mn.sortTags must be called before calling this function
// in order to sort and de-duplcate tags.
func (mn *MetricName) Marshal(dst []byte) []byte {
// Calculate the required size and pre-allocate space in dst
dstLen := len(dst)
requiredSize := len(mn.MetricGroup) + 1
for i := range mn.Tags {
tag := &mn.Tags[i]
requiredSize += len(tag.Key) + len(tag.Value) + 2
}
dst = bytesutil.ResizeWithCopy(dst, requiredSize)[:dstLen]
// Marshal MetricGroup
dst = marshalTagValue(dst, mn.MetricGroup)
// Marshal tags.
tags := mn.Tags
for i := range tags {
t := &tags[i]
dst = t.Marshal(dst)
}
return dst
}
// Unmarshal unmarshals mn from src.
func (mn *MetricName) Unmarshal(src []byte) error {
// Unmarshal MetricGroup.
var err error
src, mn.MetricGroup, err = unmarshalTagValue(mn.MetricGroup[:0], src)
if err != nil {
return fmt.Errorf("cannot unmarshal MetricGroup: %w", err)
}
mn.Tags = mn.Tags[:0]
for len(src) > 0 {
tag := mn.addNextTag()
var err error
src, err = tag.Unmarshal(src)
if err != nil {
return fmt.Errorf("cannot unmarshal tag: %w", err)
}
}
// There is no need in verifying for identical tag keys,
// since they must be handled by MetricName.sortTags before calling MetricName.Marshal.
return nil
}
// The maximum length of label name.
//
// Longer names are truncated.
const maxLabelNameLen = 256
// The maximum length of label value.
//
// Longer values are truncated.
var maxLabelValueLen = 16 * 1024
// SetMaxLabelValueLen sets the limit on the label value length.
//
// This function can be called before using the storage package.
//
// Label values with longer length are truncated.
func SetMaxLabelValueLen(n int) {
if n > 0 {
maxLabelValueLen = n
}
}
// The maximum number of labels per each timeseries.
var maxLabelsPerTimeseries = 30
// SetMaxLabelsPerTimeseries sets the limit on the number of labels
// per each time series.
//
// This function can be called before using the storage package.
//
// Superfluous labels are dropped.
func SetMaxLabelsPerTimeseries(maxLabels int) {
if maxLabels > 0 {
maxLabelsPerTimeseries = maxLabels
}
}
// MarshalMetricNameRaw marshals labels to dst and returns the result.
//
// The result must be unmarshaled with MetricName.UnmarshalRaw
func MarshalMetricNameRaw(dst []byte, labels []prompb.Label) []byte {
// Calculate the required space for dst.
dstLen := len(dst)
dstSize := dstLen
for i := range labels {
if i >= maxLabelsPerTimeseries {
trackDroppedLabels(labels, labels[i:])
break
}
label := &labels[i]
if len(label.Name) > maxLabelNameLen {
atomic.AddUint64(&TooLongLabelNames, 1)
label.Name = label.Name[:maxLabelNameLen]
}
if len(label.Value) > maxLabelValueLen {
atomic.AddUint64(&TooLongLabelValues, 1)
label.Value = label.Value[:maxLabelValueLen]
}
if len(label.Value) == 0 {
// Skip labels without values, since they have no sense in prometheus.
continue
}
if string(label.Name) == "__name__" {
label.Name = label.Name[:0]
}
dstSize += len(label.Name)
dstSize += len(label.Value)
dstSize += 4
}
dst = bytesutil.ResizeWithCopy(dst, dstSize)[:dstLen]
// Marshal labels to dst.
for i := range labels {
if i >= maxLabelsPerTimeseries {
break
}
label := &labels[i]
if len(label.Value) == 0 {
// Skip labels without values, since they have no sense in prometheus.
continue
}
dst = marshalBytesFast(dst, label.Name)
dst = marshalBytesFast(dst, label.Value)
}
return dst
}
var (
// MetricsWithDroppedLabels is the number of metrics with at least a single dropped label
MetricsWithDroppedLabels uint64
// TooLongLabelNames is the number of too long label names
TooLongLabelNames uint64
// TooLongLabelValues is the number of too long label values
TooLongLabelValues uint64
)
func trackDroppedLabels(labels, droppedLabels []prompb.Label) {
atomic.AddUint64(&MetricsWithDroppedLabels, 1)
select {
case <-droppedLabelsLogTicker.C:
// Do not call logger.WithThrottler() here, since this will result in increased CPU usage
// because labelsToString() will be called with each trackDroppedLAbels call.
logger.Warnf("dropping %d labels for %s; dropped labels: %s; either reduce the number of labels for this metric "+
"or increase -maxLabelsPerTimeseries=%d command-line flag value",
len(droppedLabels), labelsToString(labels), labelsToString(droppedLabels), maxLabelsPerTimeseries)
default:
}
}
var droppedLabelsLogTicker = time.NewTicker(5 * time.Second)
func labelsToString(labels []prompb.Label) string {
labelsCopy := append([]prompb.Label{}, labels...)
sort.Slice(labelsCopy, func(i, j int) bool {
return string(labelsCopy[i].Name) < string(labelsCopy[j].Name)
})
var b []byte
b = append(b, '{')
for i, label := range labelsCopy {
if len(label.Name) == 0 {
b = append(b, "__name__"...)
} else {
b = append(b, label.Name...)
}
b = append(b, '=')
b = strconv.AppendQuote(b, string(label.Value))
if i < len(labels)-1 {
b = append(b, ',')
}
}
b = append(b, '}')
return string(b)
}
// marshalRaw marshals mn to dst and returns the result.
//
// The results may be unmarshaled with MetricName.UnmarshalRaw.
//
// This function is for testing purposes. MarshalMetricNameRaw must be used
// in prod instead.
func (mn *MetricName) marshalRaw(dst []byte) []byte {
dst = marshalBytesFast(dst, nil)
dst = marshalBytesFast(dst, mn.MetricGroup)
mn.sortTags()
for i := range mn.Tags {
tag := &mn.Tags[i]
dst = marshalBytesFast(dst, tag.Key)
dst = marshalBytesFast(dst, tag.Value)
}
return dst
}
// UnmarshalRaw unmarshals mn encoded with MarshalMetricNameRaw.
func (mn *MetricName) UnmarshalRaw(src []byte) error {
mn.Reset()
for len(src) > 0 {
tail, key, err := unmarshalBytesFast(src)
if err != nil {
return fmt.Errorf("cannot decode key: %w", err)
}
src = tail
tail, value, err := unmarshalBytesFast(src)
if err != nil {
return fmt.Errorf("cannot decode value: %w", err)
}
src = tail
if len(key) == 0 {
mn.MetricGroup = append(mn.MetricGroup[:0], value...)
} else {
mn.AddTagBytes(key, value)
}
}
return nil
}
func marshalBytesFast(dst []byte, s []byte) []byte {
dst = encoding.MarshalUint16(dst, uint16(len(s)))
dst = append(dst, s...)
return dst
}
func unmarshalBytesFast(src []byte) ([]byte, []byte, error) {
if len(src) < 2 {
return src, nil, fmt.Errorf("cannot decode size form src=%X; it must be at least 2 bytes", src)
}
n := encoding.UnmarshalUint16(src)
src = src[2:]
if len(src) < int(n) {
return src, nil, fmt.Errorf("too short src=%X; it must be at least %d bytes", src, n)
}
return src[n:], src[:n], nil
}
// sortTags sorts tags in mn to canonical form needed for storing in the index.
//
// The function also de-duplicates tags with identical keys in mn. The last tag value
// for duplicate tags wins.
//
// Tags sorting is quite slow, so try avoiding it by caching mn
// with sorted tags.
func (mn *MetricName) sortTags() {
if len(mn.Tags) == 0 {
return
}
cts := getCanonicalTags()
if n := len(mn.Tags) - cap(cts.tags); n > 0 {
cts.tags = append(cts.tags[:cap(cts.tags)], make([]canonicalTag, n)...)
}
dst := cts.tags[:len(mn.Tags)]
for i := range mn.Tags {
tag := &mn.Tags[i]
ct := &dst[i]
ct.key = normalizeTagKey(tag.Key)
ct.tag.copyFrom(tag)
}
cts.tags = dst
// Use sort.Stable instead of sort.Sort in order to preserve the order of tags with duplicate keys.
// The last tag value wins for tags with duplicate keys.
// Use sort.Stable instead of sort.SliceStable, since sort.SliceStable allocates a lot.
sort.Stable(&cts.tags)
j := 0
var prevKey []byte
for i := range cts.tags {
tag := &cts.tags[i].tag
if j > 0 && bytes.Equal(tag.Key, prevKey) {
// Overwrite the previous tag with duplicate key.
j--
} else {
prevKey = tag.Key
}
mn.Tags[j].copyFrom(tag)
j++
}
mn.Tags = mn.Tags[:j]
putCanonicalTags(cts)
}
func getCanonicalTags() *canonicalTags {
v := canonicalTagsPool.Get()
if v == nil {
return &canonicalTags{}
}
return v.(*canonicalTags)
}
func putCanonicalTags(cts *canonicalTags) {
cts.tags = cts.tags[:0]
canonicalTagsPool.Put(cts)
}
var canonicalTagsPool sync.Pool
type canonicalTags struct {
tags canonicalTagsSort
}
type canonicalTag struct {
key []byte
tag Tag
}
type canonicalTagsSort []canonicalTag
func (ts *canonicalTagsSort) Len() int { return len(*ts) }
func (ts *canonicalTagsSort) Less(i, j int) bool {
x := *ts
return string(x[i].key) < string(x[j].key)
}
func (ts *canonicalTagsSort) Swap(i, j int) {
x := *ts
x[i], x[j] = x[j], x[i]
}
func copyTags(dst, src []Tag) []Tag {
dstLen := len(dst)
if n := dstLen + len(src) - cap(dst); n > 0 {
dst = append(dst[:cap(dst)], make([]Tag, n)...)
}
dst = dst[:dstLen+len(src)]
for i := range src {
dst[dstLen+i].copyFrom(&src[i])
}
return dst
}
var commonTagKeys = func() map[string][]byte {
lcm := map[string][]byte{
// job-like tags must go first in MetricName.Tags.
// This should improve data locality.
// They start with \x00\x00.
// Do not change values!
//
// TODO: add more job-like tags.
"namespace": []byte("\x00\x00\x00"),
"ns": []byte("\x00\x00\x01"),
"datacenter": []byte("\x00\x00\x08"),
"dc": []byte("\x00\x00\x09"),
"environment": []byte("\x00\x00\x0c"),
"env": []byte("\x00\x00\x0d"),
"cluster": []byte("\x00\x00\x10"),
"service": []byte("\x00\x00\x18"),
"job": []byte("\x00\x00\x20"),
"model": []byte("\x00\x00\x28"),
"type": []byte("\x00\x00\x30"),
"sensor_type": []byte("\x00\x00\x38"),
"SensorType": []byte("\x00\x00\x38"),
"db": []byte("\x00\x00\x40"),
// instance-like tags must go second in MetricName.Tags.
// This should improve data locality.
// They start with \x00\x01.
// Do not change values!
//
// TODO: add more instance-like tags.
"instance": []byte("\x00\x01\x00"),
"host": []byte("\x00\x01\x08"),
"server": []byte("\x00\x01\x10"),
"pod": []byte("\x00\x01\x18"),
"node": []byte("\x00\x01\x20"),
"device": []byte("\x00\x01\x28"),
"tenant": []byte("\x00\x01\x30"),
"client": []byte("\x00\x01\x38"),
"name": []byte("\x00\x01\x40"),
"measurement": []byte("\x00\x01\x48"),
}
// Generate Upper-case variants of lc
m := make(map[string][]byte, len(lcm)*2)
for k, v := range lcm {
s := strings.ToUpper(k[:1]) + k[1:]
m[k] = v
m[s] = v
}
return m
}()
func normalizeTagKey(key []byte) []byte {
tagKey := commonTagKeys[string(key)]
if tagKey == nil {
return key
}
return tagKey
}