Add batch mode to mqtt output (#4094)

etc/telegraf.conf

@@ -1791,19 +1791,19 @@
 #   ## List of metrics collected on above servers
 #   ## Each metric consists in a name, a jmx path and either
 #   ## a pass or drop slice attribute.
-#   ## This collect all heap memory usage metrics.
+#   ## This collect all heap memory usage metrics.
 #   [[inputs.jolokia.metrics]]
 #     name = "heap_memory_usage"
 #     mbean  = "java.lang:type=Memory"
 #     attribute = "HeapMemoryUsage"
 #
-#   ## This collect thread counts metrics.
+#   ## This collect thread counts metrics.
 #   [[inputs.jolokia.metrics]]
 #     name = "thread_count"
 #     mbean  = "java.lang:type=Threading"
 #     attribute = "TotalStartedThreadCount,ThreadCount,DaemonThreadCount,PeakThreadCount"
 #
-#   ## This collect number of class loaded/unloaded counts metrics.
+#   ## This collect number of class loaded/unloaded counts metrics.
 #   [[inputs.jolokia.metrics]]
 #     name = "class_count"
 #     mbean  = "java.lang:type=ClassLoading"
@@ -3561,4 +3561,3 @@
 # [[inputs.zipkin]]
 #   # path = "/api/v1/spans" # URL path for span data
 #   # port = 9411            # Port on which Telegraf listens
-

plugins/outputs/mqtt/README.md

@@ -30,6 +30,15 @@ This plugin writes to a [MQTT Broker](http://http://mqtt.org/) acting as a mqtt
   ## Use TLS but skip chain & host verification
   # insecure_skip_verify = false
 
+  ## Batch messages in a topic
+  ## batch = false
+  ## Flag to determine if messages sent in a topic in a flush interval,
+  ## need to be batched into one message.
+  ## batch = true, batches the messages in a topic to one messages
+  ## batch = false, default behaviour
+  # batch = false
+
+
   ## Data format to output.
   data_format = "influx"
 ```

plugins/outputs/mqtt/mqtt.go

@@ -45,6 +45,15 @@ var sampleConfig = `
   ## more about them here:
   ## https://github.com/influxdata/telegraf/blob/master/docs/DATA_FORMATS_OUTPUT.md
   data_format = "influx"
+
+  ## Batch messages in a topic
+  ## batch = false
+  ## Flag to determine if messages sent in a topic in a flush interval,
+  ## need to be batched into one message.
+  ## batch = true, batches the messages in a topic to one messages
+  ## batch = false, default behaviour
+  # batch = false
+
 `
 
 type MQTT struct {
@@ -57,6 +66,7 @@ type MQTT struct {
 	QoS         int    `toml:"qos"`
 	ClientID    string `toml:"client_id"`
 	tls.ClientConfig
+	BatchMessage bool `toml:"batch"`
 
 	client paho.Client
 	opts   *paho.ClientOptions
@@ -117,6 +127,8 @@ func (m *MQTT) Write(metrics []telegraf.Metric) error {
 		hostname = ""
 	}
 
+	metricsmap := make(map[string][]telegraf.Metric)
+
 	for _, metric := range metrics {
 		var t []string
 		if m.TopicPrefix != "" {
@@ -129,7 +141,11 @@ func (m *MQTT) Write(metrics []telegraf.Metric) error {
 		t = append(t, metric.Name())
 		topic := strings.Join(t, "/")
 
+		if m.BatchMessage {
+			metricsmap[topic] = append(metricsmap[topic], metric)
+		} else {
 			buf, err := m.serializer.Serialize(metric)
+
 			if err != nil {
 				return err
 			}
@@ -139,6 +155,19 @@ func (m *MQTT) Write(metrics []telegraf.Metric) error {
 				return fmt.Errorf("Could not write to MQTT server, %s", err)
 			}
 		}
+	}
+
+	for key := range metricsmap {
+		buf, err := m.serializer.SerializeBatch(metricsmap[key])
+
+		if err != nil {
+			return err
+		}
+		publisherr := m.publish(key, buf)
+		if publisherr != nil {
+			return fmt.Errorf("Could not write to MQTT server, %s", publisherr)
+		}
+	}
 
 	return nil
 }