The following instrumentation types are available:

• Automatic instrumentation for Kafka-based workloads
• Manual Instrumentation for Kafka-based workloads
• Manual instrumentation for other queuing technology or protocol

Prerequisites

To start with Data Streams Monitoring, you need recent versions of the Datadog Agent and Data Streams Monitoring libraries:

• Datadog Agent v7.34.0 or later
• dd-trace-go v1.56.1 or later

Supported libraries

Installation

Monitoring Kafka Pipelines

Data Streams Monitoring uses message headers to propagate context through Kafka streams. If log.message.format.version is set in the Kafka broker configuration, it must be set to 0.11.0.0 or higher. Data Streams Monitoring is not supported for versions lower than this.

Monitoring RabbitMQ pipelines

The RabbitMQ integration can provide detailed monitoring and metrics of your RabbitMQ deployments. For full compatibility with Data Streams Monitoring, Datadog recommends configuring the integration as follows:

instances:
  - prometheus_plugin:
      url: http://<HOST>:15692
      unaggregated_endpoint: detailed?family=queue_coarse_metrics&family=queue_consumer_count&family=channel_exchange_metrics&family=channel_queue_exchange_metrics&family=node_coarse_metrics

This ensures that all RabbitMQ graphs populate, and that you see detailed metrics for individual exchanges as well as queues.

Automatic Instrumentation

Automatic instrumentation uses Orchestrion to install dd-trace-go and supports both the Sarama and Confluent Kafka libraries.

To automatically instrument your service:

1. Follow the Orchestrion Getting Started guide to compile or run your service using Orchestrion.
2. Set the DD_DATA_STREAMS_ENABLED=true environment variable

Manual instrumentation

Sarama Kafka client

To manually instrument the Sarama Kafka client with Data Streams Monitoring:

1. Import the ddsarama go library

import (
  ddsarama "gopkg.in/DataDog/dd-trace-go.v1/contrib/Shopify/sarama" // 1.x
  // ddsarama "github.com/DataDog/dd-trace-go/contrib/Shopify/sarama/v2" // 2.x
)

2. Wrap the producer with `ddsarama.WrapAsyncProducer`

...
config := sarama.NewConfig()
producer, err := sarama.NewAsyncProducer([]string{bootStrapServers}, config)

// ADD THIS LINE
producer = ddsarama.WrapAsyncProducer(config, producer, ddsarama.WithDataStreams())

Confluent Kafka client

To manually instrument Confluent Kafka with Data Streams Monitoring:

1. Import the ddkafka go library

import (
  ddkafka "gopkg.in/DataDog/dd-trace-go.v1/contrib/confluentinc/confluent-kafka-go/kafka.v2" // 1.x
  // ddkafka "github.com/DataDog/dd-trace-go/contrib/confluentinc/confluent-kafka-go/kafka.v2/v2" // 2.x
)

2. Wrap the producer creation with ddkafka.NewProducer and use the ddkafka.WithDataStreams() configuration

// CREATE PRODUCER WITH THIS WRAPPER
producer, err := ddkafka.NewProducer(&kafka.ConfigMap{
		"bootstrap.servers": bootStrapServers,
}, ddkafka.WithDataStreams())

If a service consumes data from one point and produces to another point, propagate context between the two places using the Go context structure:

3. Extract the context from headers

ctx = datastreams.ExtractFromBase64Carrier(ctx, ddsarama.NewConsumerMessageCarrier(message))

4. Inject it into the header before producing downstream

   datastreams.InjectToBase64Carrier(ctx, ddsarama.NewProducerMessageCarrier(message))
   

Other queuing technologies or protocols

You can also use manual instrumentation. For example, you can propagate context through Kinesis.

Instrumenting the produce call

1. Ensure your message supports the TextMapWriter interface.
2. Inject the context into your message and instrument the produce call by calling:

ctx, ok := tracer.SetDataStreamsCheckpointWithParams(ctx, options.CheckpointParams{PayloadSize: getProducerMsgSize(msg)}, "direction:out", "type:kinesis", "topic:kinesis_arn")
if ok {
  datastreams.InjectToBase64Carrier(ctx, message)
}

Instrumenting the consume call

1. Ensure your message supports the TextMapReader interface.
2. Extract the context from your message and instrument the consume call by calling:

	ctx, ok := tracer.SetDataStreamsCheckpointWithParams(datastreams.ExtractFromBase64Carrier(context.Background(), message), options.CheckpointParams{PayloadSize: payloadSize}, "direction:in", "type:kinesis", "topic:kinesis_arn")

Monitoring connectors

Confluent Cloud connectors

Data Streams Monitoring can automatically discover your Confluent Cloud connectors and visualize them within the context of your end-to-end streaming data pipeline.

Setup

1. Install and configure the Datadog-Confluent Cloud integration.

2. In Datadog, open the Confluent Cloud integration tile.

   

   Under Actions, a list of resources populates with detected clusters and connectors. Datadog attempts to discover new connectors every time you view this integration tile.

3. Select the resources you want to add.

4. Click Add Resources.

5. Navigate to Data Streams Monitoring to visualize the connectors and track connector status and throughput.

Further reading

Documentation, liens et articles supplémentaires utiles:

Autodiscover Confluent Cloud connectors and easily monitor performance in Data Streams MonitoringBLOG