Initial version of MQTT sink

sinks/amqpSink.md

@@ -1,33 +0,0 @@
-## `amqp` sink
-
-The `amqp` sink publishes all metrics into a RabbitMQ network. The publishing key is the queue name in the configuration file
-
-### Configuration structure
-
-```json
-{
-  "<name>": {
-    "type": "amqp",
-    "queue_name" : "myqueue",
-    "batch_size" : 1000,
-    "flush_delay": "4s",
-    "publish_timeout": "1s",
-    "host": "dbhost.example.com",
-    "port": 5672,
-    "username": "exampleuser",
-    "password" : "examplepw",
-    "meta_as_tags" : [],
-  }
-}
-```
-
-- `type`: makes the sink an `amqp` sink, also `rabbitmq` is allowed as alias
-- `queue_name`: All metrics are published to this queue
-- `host`: Hostname of the RabbitMQ server
-- `port`: Port number of the RabbitMQ server
-- `username`: Username for basic authentication
-- `password`: Password for basic authentication
-- `meta_as_tags`: print all meta information as tags in the output (optional)
-- `publish_timeout`: Timeout for each publication operation (default `1s`)
-- `flush_delay`: Group metrics coming in to a single batch (default `4s`)
-- `batch_size`: Maximal batch size. If `batch_size` is reached before the end of `flush_delay`, the metrics are sent without further delay (default: `1000`)

sinks/mqttSink.go

@@ -2,26 +2,28 @@ package sinks
 
 import (
 	"bytes"
-	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"net"
+	"os"
 	"sync"
 	"time"
 
 	cclog "github.com/ClusterCockpit/cc-metric-collector/pkg/ccLogger"
 	lp "github.com/ClusterCockpit/cc-metric-collector/pkg/ccMetric"
+	"github.com/go-mqtt/mqtt"
 	influx "github.com/influxdata/line-protocol/v2/lineprotocol"
-	amqp "github.com/rabbitmq/amqp091-go"
 	"golang.org/x/exp/slices"
 )
 
-type AmqpSinkConfig struct {
+type MqttSinkConfig struct {
 	// defines JSON tags for 'type' and 'meta_as_tags' (string list)
 	// See: metricSink.go
 	defaultSinkConfig
-	// Additional config options, for AmqpSink
-	QueueName string `json:"queue_name"`
+	// Additional config options, for MqttSink
+	ClientID             string `json:"client_id"`
+	PersistenceDirectory string `json:"persistence_directory,omitempty"`
 	// Maximum number of points sent to server in single request.
 	// Default: 1000
 	BatchSize int `json:"batch_size,omitempty"`
@@ -33,18 +35,20 @@ type AmqpSinkConfig struct {
 	FlushInterval string `json:"flush_delay,omitempty"`
 	flushDelay    time.Duration
 
-	Hostname       string `json:"hostname"`
-	Port           int    `json:"port"`
-	PublishTimeout string `json:"publish_timeout,omitempty"`
-	publishTimeout time.Duration
-	Username       string `json:"username,omitempty"`
-	Password       string `json:"password,omitempty"`
+	DialProtocol string `json:"dial_protocol"`
+	Hostname     string `json:"hostname"`
+	Port         int    `json:"port"`
+	PauseTimeout string `json:"pause_timeout"`
+	pauseTimeout time.Duration
+	KeepAlive    uint16 `json:"keep_alive_seconds"`
+	Username     string `json:"username,omitempty"`
+	Password     string `json:"password,omitempty"`
 }
 
-type AmqpSink struct {
+type MqttSink struct {
 	// declares elements 	'name' and 'meta_as_tags' (string to bool map!)
 	sink
-	config AmqpSinkConfig // entry point to the AmqpSinkConfig
+	config MqttSinkConfig // entry point to the MqttSinkConfig
 	// influx line protocol encoder
 	encoder influx.Encoder
 	// number of records stored in the encoder
@@ -63,9 +67,8 @@ type AmqpSink struct {
 	// WaitGroup to ensure only one send operation is running at a time
 	sendWaitGroup sync.WaitGroup
 
-	client  *amqp.Connection
-	channel *amqp.Channel
-	queue   amqp.Queue
+	client     *mqtt.Client
+	mqttconfig mqtt.Config
 }
 
 // Implement functions required for Sink interface
@@ -73,7 +76,7 @@ type AmqpSink struct {
 // See: metricSink.go
 
 // Code to submit a single CCMetric to the sink
-func (s *AmqpSink) Write(m lp.CCMetric) error {
+func (s *MqttSink) Write(m lp.CCMetric) error {
 
 	// Lock for encoder usage
 	s.encoderLock.Lock()
@@ -194,7 +197,7 @@ func (s *AmqpSink) Write(m lp.CCMetric) error {
 }
 
 // If the sink uses batched sends internally, you can tell to flush its buffers
-func (s *AmqpSink) Flush() error {
+func (s *MqttSink) Flush() error {
 
 	// Lock for encoder usage
 	// Own lock for as short as possible: the time it takes to clone the buffer.
@@ -219,21 +222,37 @@ func (s *AmqpSink) Flush() error {
 	go func() {
 		defer s.sendWaitGroup.Done()
 		//startTime := time.Now()
-		ctx, cancel := context.WithTimeout(context.Background(), s.config.flushDelay)
-		defer cancel()
-		err := s.channel.PublishWithContext(ctx, "", s.queue.Name, false, false, amqp.Publishing{
-			ContentType: "text/plain",
-			Body:        buf,
-		})
-		if err != nil {
-			cclog.ComponentError(s.name, err.Error())
+		for {
+			exchange, err := s.client.PublishAtLeastOnce(buf, s.config.ClientID)
+			switch {
+			case err == nil:
+				return
+
+			case mqtt.IsDeny(err), errors.Is(err, mqtt.ErrClosed):
+				return
+
+			case errors.Is(err, mqtt.ErrMax):
+				time.Sleep(s.config.pauseTimeout)
+
+			default:
+				time.Sleep(s.config.pauseTimeout)
+				continue
+			}
+
+			for err := range exchange {
+				if errors.Is(err, mqtt.ErrClosed) {
+					return
+				}
+
+			}
+			return
 		}
 	}()
 	return nil
 }
 
 // Close sink: close network connection, close files, close libraries, ...
-func (s *AmqpSink) Close() {
+func (s *MqttSink) Close() {
 
 	cclog.ComponentDebug(s.name, "CLOSE")
 
@@ -258,18 +277,19 @@ func (s *AmqpSink) Close() {
 
 // New function to create a new instance of the sink
 // Initialize the sink by giving it a name and reading in the config JSON
-func NewAmqpSink(name string, config json.RawMessage) (Sink, error) {
-	s := new(AmqpSink)
+func NewMqttSink(name string, config json.RawMessage) (Sink, error) {
+	s := new(MqttSink)
 
 	// Set name of sampleSink
-	// The name should be chosen in such a way that different instances of AmqpSink can be distinguished
-	s.name = fmt.Sprintf("AmqpSink(%s)", name) // Always specify a name here
+	// The name should be chosen in such a way that different instances of MqttSink can be distinguished
+	s.name = fmt.Sprintf("MqttSink(%s)", name) // Always specify a name here
 
 	// Set defaults in s.config
 	// Allow overwriting these defaults by reading config JSON
 
-	s.config.PublishTimeout = "4s"
-	s.config.publishTimeout = time.Duration(4) * time.Second
+	s.config.PauseTimeout = "4s"
+	s.config.pauseTimeout = time.Duration(4) * time.Second
+	s.config.DialProtocol = "tcp"
 	s.config.Hostname = "localhost"
 	s.config.Port = 1883
 
@@ -291,10 +311,10 @@ func NewAmqpSink(name string, config json.RawMessage) (Sink, error) {
 
 	// Check if all required fields in the config are set
 	// E.g. use 'len(s.config.Option) > 0' for string settings
-	if t, err := time.ParseDuration(s.config.PublishTimeout); err == nil {
-		s.config.publishTimeout = t
+	if t, err := time.ParseDuration(s.config.PauseTimeout); err == nil {
+		s.config.pauseTimeout = t
 	} else {
-		err := fmt.Errorf("to parse duration for PublishTimeout: %s", s.config.PublishTimeout)
+		err := fmt.Errorf("to parse duration for PauseTimeout: %s", s.config.PauseTimeout)
 		cclog.ComponentError(s.name, err.Error())
 		return nil, err
 	}
@@ -306,46 +326,46 @@ func NewAmqpSink(name string, config json.RawMessage) (Sink, error) {
 		return nil, err
 	}
 
-	url := net.JoinHostPort(s.config.Hostname, fmt.Sprintf("%d", s.config.Port))
-	userpart := ""
-	if len(s.config.Username) > 0 {
-		userpart = s.config.Username
-		if len(s.config.Password) > 0 {
-			userpart += ":" + s.config.Password
-		}
-		userpart += "@"
+	switch s.config.DialProtocol {
+	case "tcp", "udp":
+	default:
+		err := errors.New("failed to parse dial protocol, allowed: tcp, udp")
+		cclog.ComponentError(s.name, err.Error())
+		return nil, err
+	}
+
+	var persistence mqtt.Persistence
+	if len(s.config.PersistenceDirectory) > 0 {
+		persistence = mqtt.FileSystem(s.config.PersistenceDirectory)
+	} else {
+		tmpdir, err := os.MkdirTemp("", "mqtt")
+		if err == nil {
+			persistence = mqtt.FileSystem(tmpdir)
+		}
 	}
-	url = fmt.Sprintf("amqp://%s%s", userpart, url)
 
 	// Establish connection to the server, library, ...
 	// Check required files exist and lookup path(s) of executable(s)
-	c, err := amqp.Dial(url)
-	if err != nil {
-		return nil, err
-	}
-	s.client = c
 
-	ch, err := c.Channel()
-	if err != nil {
-		s.client.Close()
-		return nil, err
-	}
-	s.channel = ch
+	dialer := mqtt.NewDialer(s.config.DialProtocol, net.JoinHostPort(s.config.Hostname, fmt.Sprintf("%d", s.config.Port)))
 
-	q, err := ch.QueueDeclare(
-		s.config.QueueName, // name
-		false,              // durable
-		false,              // delete when unused
-		false,              // exclusive
-		false,              // no-wait
-		nil,                // arguments
-	)
+	s.mqttconfig = mqtt.Config{
+		Dialer:       dialer,
+		PauseTimeout: s.config.pauseTimeout,
+		KeepAlive:    uint16(s.config.KeepAlive),
+	}
+	if len(s.config.Username) > 0 {
+		s.mqttconfig.UserName = s.config.Username
+	}
+	if len(s.config.Password) > 0 {
+		s.mqttconfig.Password = []byte(s.config.Password)
+	}
+
+	client, err := mqtt.InitSession(s.config.ClientID, persistence, &s.mqttconfig)
 	if err != nil {
-		s.channel.Close()
-		s.client.Close()
 		return nil, err
 	}
-	s.queue = q
+	s.client = client
 
 	// Return (nil, meaningful error message) in case of errors
 	return s, nil

sinks/mqttSink.md

@@ -0,0 +1,39 @@
+## `mqtt` sink
+
+The `mqtt` sink publishes all metrics into a MQTT network.
+
+### Configuration structure
+
+```json
+{
+  "<name>": {
+    "type": "mqtt",
+    "client_id" : "myid",
+    "persistence_directory": "/tmp",
+    "batch_size": 1000,
+    "flush_delay": "1s",
+    "dial_protocol": "tcp",
+    "host": "dbhost.example.com",
+    "port": 1883,
+    "user": "exampleuser",
+    "password" : "examplepw",
+    "pause_timeout": "1s",
+    "keep_alive_seconds": 10,
+    "meta_as_tags" : [],
+  }
+}
+```
+
+- `type`: makes the sink an `mqtt` sink
+- `client_id`: MQTT clients use a client_id to talk to the broker
+- `persistence_directory`: MQTT stores messages temporarly on disc if the broker is not available. Folder needs to be writable (default: `/tmp`)
+- `pause_timeout`: Waittime when published failed
+- `keep_alive_seconds`: Keep the connection alive for some time. Recommended to be longer than global `interval`.
+- `flush_delay`: Group metrics coming in to a single batch
+- `batch_size`: Maximal batch size. If `batch_size` is reached before the end of `flush_delay`, the metrics are sent without further delay
+- `dial_protocol`: Use `tcp` or `udp` for the MQTT communication
+- `host`: Hostname of the MQTT broker
+- `port`: Port number of the MQTT broker
+- `user`: Username for authentication
+- `password`: Password for authentication
+- `meta_as_tags`: print all meta information as tags in the output (optional)

sinks/sinkManager.go

@@ -21,8 +21,7 @@ var AvailableSinks = map[string]func(name string, config json.RawMessage) (Sink,
 	"influxdb":    NewInfluxSink,
 	"influxasync": NewInfluxAsyncSink,
 	"http":        NewHttpSink,
-	"amqp":        NewAmqpSink,
-	"rabbitmq":    NewAmqpSink,
+	"mqtt":        NewMqttSink,
 }
 
 // Metric collector manager data structure