Merge pull request #6743 from influxdata/jw-parse-key

Optimize series key parsing on startup

CHANGELOG.md

@@ -50,6 +50,7 @@
 - [#6599](https://github.com/influxdata/influxdb/issues/6599): Ensure that future points considered in SHOW queries.
 - [#6720](https://github.com/influxdata/influxdb/issues/6720): Concurrent map read write panic. Thanks @arussellsaw
 - [#6727](https://github.com/influxdata/influxdb/issues/6727): queries with strings that look like dates end up with date types, not string types
+- [#6250](https://github.com/influxdata/influxdb/issues/6250): Slow startup time
 
 ## v0.13.0 [2016-05-12]
 

models/points.go

@@ -1040,6 +1040,10 @@ func escapeTag(in []byte) []byte {
 }
 
 func unescapeTag(in []byte) []byte {
+	if bytes.IndexByte(in, '\\') == -1 {
+		return in
+	}
+
 	for b, esc := range tagEscapeCodes {
 		if bytes.IndexByte(in, b) != -1 {
 			in = bytes.Replace(in, esc, []byte{b}, -1)
@@ -1216,7 +1220,8 @@ func (p *point) Tags() Tags {
 }
 
 func parseTags(buf []byte) Tags {
-	tags := map[string]string{}
+	tags := make(map[string]string, bytes.Count(buf, []byte(",")))
+	hasEscape := bytes.IndexByte(buf, '\\') != -1
 
 	if len(buf) != 0 {
 		pos, name := scanTo(buf, 0, ',')
@@ -1239,7 +1244,11 @@ func parseTags(buf []byte) Tags {
 				continue
 			}
 
-			tags[string(unescapeTag(key))] = string(unescapeTag(value))
+			if hasEscape {
+				tags[string(unescapeTag(key))] = string(unescapeTag(value))
+			} else {
+				tags[string(key)] = string(value)
+			}
 
 			i++
 		}

models/points_test.go

@@ -112,6 +112,13 @@ func BenchmarkParsePointsTagsUnSorted10(b *testing.B) {
 	}
 }
 
+func BenchmarkParseKey(b *testing.B) {
+	line := `cpu,region=us-west,host=serverA,env=prod,target=servers,zone=1c,tag1=value1,tag2=value2,tag3=value3,tag4=value4,tag5=value5`
+	for i := 0; i < b.N; i++ {
+		models.ParseKey(line)
+	}
+}
+
 // TestPoint wraps a models.Point but also makes available the raw
 // arguments to the Point.
 //

pkg/escape/strings.go

@@ -20,6 +20,10 @@ func init() {
 }
 
 func UnescapeString(in string) string {
+	if strings.IndexByte(in, '\\') == -1 {
+		return in
+	}
+
 	for b, esc := range codesStr {
 		in = strings.Replace(in, esc, b, -1)
 	}

tsdb/meta.go

@@ -133,14 +133,13 @@ func (d *DatabaseIndex) CreateSeriesIndexIfNotExists(measurementName string, ser
 	m := d.CreateMeasurementIndexIfNotExists(measurementName)
 
 	d.mu.Lock()
-	defer d.mu.Unlock()
-
 	// set the in memory ID for query processing on this shard
 	series.id = d.lastID + 1
 	d.lastID++
 
 	series.measurement = m
 	d.series[series.Key] = series
+	d.mu.Unlock()
 
 	m.AddSeries(series)
 
@@ -502,8 +501,7 @@ type Measurement struct {
 	fieldNames map[string]struct{}
 
 	// in-memory index fields
-	seriesByID          map[uint64]*Series // lookup table for series by their id
-	measurement         *Measurement
+	seriesByID          map[uint64]*Series              // lookup table for series by their id
 	seriesByTagKeyValue map[string]map[string]SeriesIDs // map from tag key to value to sorted set of series ids
 	seriesIDs           SeriesIDs                       // sorted list of series IDs in this measurement
 }
@@ -516,7 +514,7 @@ func NewMeasurement(name string) *Measurement {
 
 		seriesByID:          make(map[uint64]*Series),
 		seriesByTagKeyValue: make(map[string]map[string]SeriesIDs),
-		seriesIDs:           make(SeriesIDs, 0),
+		seriesIDs:           make(SeriesIDs, 0, 1),
 	}
 }
 
@@ -591,12 +589,20 @@ func (m *Measurement) HasSeries() bool {
 
 // AddSeries will add a series to the measurementIndex. Returns false if already present
 func (m *Measurement) AddSeries(s *Series) bool {
+	m.mu.RLock()
+	if _, ok := m.seriesByID[s.id]; ok {
+		m.mu.RUnlock()
+		return false
+	}
+	m.mu.RUnlock()
+
 	m.mu.Lock()
 	defer m.mu.Unlock()
 
 	if _, ok := m.seriesByID[s.id]; ok {
 		return false
 	}
+
 	m.seriesByID[s.id] = s
 	m.seriesIDs = append(m.seriesIDs, s.id)