Merge pull request #267 from twooster/add-librato-tag-support

Add librato tag support

README.md

@@ -29,15 +29,48 @@ We recommend this only as an intermediate solution and recommend switching to
 [native Prometheus instrumentation](http://prometheus.io/docs/instrumenting/clientlibs/)
 in the long term.
 
-### DogStatsD extensions
+### Tagging Extensions
 
-The exporter will convert DogStatsD-style tags to prometheus labels. See
-[Tags](https://docs.datadoghq.com/developers/dogstatsd/data_types/#tagging) in the DogStatsD
-documentation for the concept description and
-[Datagram Format](https://docs.datadoghq.com/developers/dogstatsd/datagram_shell/)
-for specifics. It boils down to appending
-`|#tag:value,another_tag:another_value` to the normal StatsD format.  Tags
-without values (`#some_tag`) are not supported.
+The exporter supports Librato, InfluxDB, and DogStatsD-style tags,
+which will be converted into Prometheus labels.
+
+For Librato-style tags, they must be appended to the metric name with a
+delimiting `#`, as so:
+
+```
+metric.name#tagName=val,tag2Name=val2:0|c
+```
+
+See the [statsd-librato-backend README](https://github.com/librato/statsd-librato-backend#tags)
+for a more complete description.
+
+For InfluxDB-style tags, they must be appended to the metric name with a
+delimiting comma, as so:
+
+```
+metric.name,tagName=val,tag2Name=val2:0|c
+```
+
+See [this InfluxDB blog post](https://www.influxdata.com/blog/getting-started-with-sending-statsd-metrics-to-telegraf-influxdb/#introducing-influx-statsd)
+for a larger overview.
+
+
+For DogStatsD-style tags, they're appended as a `|#` delimited section at the
+end of the metric, as so:
+
+```
+metric.name:0|c|#tagName=val,tag2Name=val2
+```
+
+See [Tags](https://docs.datadoghq.com/developers/dogstatsd/data_types/#tagging)
+in the DogStatsD documentation for the concept description and
+[Datagram Format](https://docs.datadoghq.com/developers/dogstatsd/datagram_shell/).
+If you encounter problems, note that this tagging style is incompatible with
+the original `statsd` implementation.
+
+Be aware: If you mix tag styles (e.g., Librato/InfluxDB with DogStatsD), the
+exporter will consider this an error and the sample will be discarded. Also,
+tags without values (`#some_tag`) are not supported and will be ignored.
 
 ## Building and Running
 

bridge_test.go

@@ -127,6 +127,46 @@ func TestHandlePacket(t *testing.T) {
 					labels:     map[string]string{"tag1": "bar", "tag2": "baz"},
 				},
 			},
+		}, {
+			name: "librato tag extension",
+			in:   "foo#tag1=bar,tag2=baz:100|c",
+			out: Events{
+				&CounterEvent{
+					metricName: "foo",
+					value:      100,
+					labels:     map[string]string{"tag1": "bar", "tag2": "baz"},
+				},
+			},
+		}, {
+			name: "librato tag extension with tag keys unsupported by prometheus",
+			in:   "foo#09digits=0,tag.with.dots=1:100|c",
+			out: Events{
+				&CounterEvent{
+					metricName: "foo",
+					value:      100,
+					labels:     map[string]string{"_09digits": "0", "tag_with_dots": "1"},
+				},
+			},
+		}, {
+			name: "influxdb tag extension",
+			in:   "foo,tag1=bar,tag2=baz:100|c",
+			out: Events{
+				&CounterEvent{
+					metricName: "foo",
+					value:      100,
+					labels:     map[string]string{"tag1": "bar", "tag2": "baz"},
+				},
+			},
+		}, {
+			name: "influxdb tag extension with tag keys unsupported by prometheus",
+			in:   "foo,09digits=0,tag.with.dots=1:100|c",
+			out: Events{
+				&CounterEvent{
+					metricName: "foo",
+					value:      100,
+					labels:     map[string]string{"_09digits": "0", "tag_with_dots": "1"},
+				},
+			},
 		}, {
 			name: "datadog tag extension",
 			in:   "foo:100|c|#tag1:bar,tag2:baz",
@@ -197,6 +237,18 @@ func TestHandlePacket(t *testing.T) {
 					labels:     map[string]string{"tag1": "bar", "tag2": "baz"},
 				},
 			},
+		}, {
+			name: "librato/dogstatsd mixed tag styles without sampling",
+			in:   "foo#tag1=foo,tag3=bing:100|c|#tag1:bar,#tag2:baz",
+			out:  Events{},
+		}, {
+			name: "influxdb/dogstatsd mixed tag styles without sampling",
+			in:   "foo,tag1=foo,tag3=bing:100|c|#tag1:bar,#tag2:baz",
+			out:  Events{},
+		}, {
+			name: "mixed tag styles with sampling",
+			in:   "foo#tag1=foo,tag3=bing:100|c|@0.1|#tag1:bar,#tag2:baz",
+			out:  Events{},
 		}, {
 			name: "histogram with sampling",
 			in:   "foo:0.01|h|@0.2|#tag1:bar,#tag2:baz",

exporter.go

@@ -275,60 +275,88 @@ func buildEvent(statType, metric string, value float64, relative bool, labels ma
 	}
 }
 
-func handleDogStatsDTagToKeyValue(labels map[string]string, component, tag string) {
-	// Bail early if the tag is empty
+func parseTag(component, tag string, separator rune, labels map[string]string) {
+	// Entirely empty tag is an error
 	if len(tag) == 0 {
 		tagErrors.Inc()
-		log.Debugf("Malformed or empty DogStatsD tag %s in component %s", tag, component)
+		log.Debugf("Empty name tag in component %s", component)
 		return
 	}
-	// Skip hash if found.
-	if tag[0] == '#' {
-		tag = tag[1:]
-	}
 
-	// find the first comma and split the tag into key and value.
-	var k, v string
 	for i, c := range tag {
-		if c == ':' {
-			k = tag[0:i]
-			v = tag[(i + 1):]
-			break
-		}
-	}
-	// If either of them is empty, then either the k or v is empty, or we
-	// didn't find a colon, either way, throw an error and skip ahead.
+		if c == separator {
+			k := tag[:i]
+			v := tag[i+1:]
+
 			if len(k) == 0 || len(v) == 0 {
+				// Empty key or value is an error
 				tagErrors.Inc()
-		log.Debugf("Malformed or empty DogStatsD tag %s in component %s", tag, component)
+				log.Debugf("Malformed name tag %s=%s in component %s", k, v, component)
+			} else {
+				labels[escapeMetricName(k)] = v
+			}
 			return
 		}
+	}
 
-	labels[escapeMetricName(k)] = v
-
-	return
+	// Missing separator (no value) is an error
+	tagErrors.Inc()
+	log.Debugf("Malformed name tag %s in component %s", tag, component)
 }
 
-func parseDogStatsDTagsToLabels(component string) map[string]string {
-	labels := map[string]string{}
-	tagsReceived.Inc()
-
+func parseNameTags(component string, labels map[string]string) {
 	lastTagEndIndex := 0
 	for i, c := range component {
 		if c == ',' {
 			tag := component[lastTagEndIndex:i]
 			lastTagEndIndex = i + 1
-			handleDogStatsDTagToKeyValue(labels, component, tag)
+			parseTag(component, tag, '=', labels)
 		}
 	}
 
 	// If we're not off the end of the string, add the last tag
 	if lastTagEndIndex < len(component) {
 		tag := component[lastTagEndIndex:]
-		handleDogStatsDTagToKeyValue(labels, component, tag)
+		parseTag(component, tag, '=', labels)
+	}
+}
+
+func trimLeftHash(s string) string {
+	if s != "" && s[0] == '#' {
+		return s[1:]
+	}
+	return s
+}
+
+func parseDogStatsDTags(component string, labels map[string]string) {
+	lastTagEndIndex := 0
+	for i, c := range component {
+		if c == ',' {
+			tag := component[lastTagEndIndex:i]
+			lastTagEndIndex = i + 1
+			parseTag(component, trimLeftHash(tag), ':', labels)
+		}
 	}
 
-	return labels
+	// If we're not off the end of the string, add the last tag
+	if lastTagEndIndex < len(component) {
+		tag := component[lastTagEndIndex:]
+		parseTag(component, trimLeftHash(tag), ':', labels)
+	}
+}
+
+func parseNameAndTags(name string, labels map[string]string) string {
+	for i, c := range name {
+		// `#` delimits start of tags by Librato
+		// https://www.librato.com/docs/kb/collect/collection_agents/stastd/#stat-level-tags
+		// `,` delimits start of tags by InfluxDB
+		// https://www.influxdata.com/blog/getting-started-with-sending-statsd-metrics-to-telegraf-influxdb/#introducing-influx-statsd
+		if c == '#' || c == ',' {
+			parseNameTags(name[i+1:], labels)
+			return name[:i]
+		}
+	}
+	return name
 }
 
 func lineToEvents(line string) Events {
@@ -343,10 +371,22 @@ func lineToEvents(line string) Events {
 		log.Debugln("Bad line from StatsD:", line)
 		return events
 	}
-	metric := elements[0]
+
+	labels := map[string]string{}
+	metric := parseNameAndTags(elements[0], labels)
+
 	var samples []string
 	if strings.Contains(elements[1], "|#") {
-		// using datadog extensions, disable multi-metrics
+		// using DogStatsD tags
+
+		// don't allow mixed tagging styles
+		if len(labels) > 0 {
+			sampleErrors.WithLabelValues("mixed_tagging_styles").Inc()
+			log.Debugln("Bad line (multiple tagging styles) from StatsD:", line)
+			return events
+		}
+
+		// disable multi-metrics
 		samples = elements[1:]
 	} else {
 		samples = strings.Split(elements[1], ":")
@@ -376,7 +416,6 @@ samples:
 		}
 
 		multiplyEvents := 1
-		labels := map[string]string{}
 		if len(components) >= 3 {
 			for _, component := range components[2:] {
 				if len(component) == 0 {
@@ -407,7 +446,7 @@ samples:
 						multiplyEvents = int(1 / samplingFactor)
 					}
 				case '#':
-					labels = parseDogStatsDTagsToLabels(component[1:])
+					parseDogStatsDTags(component[1:], labels)
 				default:
 					log.Debugf("Invalid sampling factor or tag section %s on line %s", components[2], line)
 					sampleErrors.WithLabelValues("invalid_sample_factor").Inc()
@@ -416,6 +455,10 @@ samples:
 			}
 		}
 
+		if len(labels) > 0 {
+			tagsReceived.Inc()
+		}
+
 		for i := 0; i < multiplyEvents; i++ {
 			event, err := buildEvent(statType, metric, value, relative, labels)
 			if err != nil {

exporter_test.go

@@ -941,7 +941,7 @@ func BenchmarkEscapeMetricName(b *testing.B) {
 	}
 }
 
-func BenchmarkParseDogStatsDTagsToLabels(b *testing.B) {
+func BenchmarkParseDogStatsDTags(b *testing.B) {
 	scenarios := map[string]string{
 		"1 tag w/hash":         "#test:tag",
 		"1 tag w/o hash":       "test:tag",
@@ -953,7 +953,8 @@ func BenchmarkParseDogStatsDTagsToLabels(b *testing.B) {
 	for name, tags := range scenarios {
 		b.Run(name, func(b *testing.B) {
 			for n := 0; n < b.N; n++ {
-				parseDogStatsDTagsToLabels(tags)
+				labels := map[string]string{}
+				parseDogStatsDTags(tags, labels)
 			}
 		})
 	}