Building a Connector

Connectors in OpenTelemetry

The content of this page is most applicable if you already have an instrumented application generating some kind of tracing telemetry data and already have an understanding of the OpenTelemetry Collector.

What is a Connector?

A connector acts as the means of sending telemetry data between different collector pipelines by connecting them. A connector acts as an exporter to one pipeline and a receiver to another. Each pipeline in the OpenTelemetry Collector acts on one type of telemetry data. There may exist the need to process one form of telemetry data into another one, but it is required to route the according data to its proper collector pipeline.

Why use a Connector?

The connector is beneficial at merging, routing and replicating data streams. Along with sequential pipelining, which is to connect pipelines together, the connector component is capable of conditional data flow and generated data streams. Conditional data flow means sending data to the highest priority pipeline and has error detection to route to alternative pipeline if need be. Generated data streams means that the component generates and emits its own data based on the received data. This tutorial emphasizes on the connector’s ability to connect pipelines.

There are processors in OpenTelemetry that convert telemetry data of one type into another one. A few examples are the spanmetrics processor, as well as the servicegraph processor. The spanmetrics processor generates aggregate requests, error and duration metrics from span data. T​he servicegraph processor analyzes trace data and generates metrics that describe the relationship between the services. Both these processors ingest trace data and convert them to metrics data. Since pipelines in the OpenTelemetry Collector are for only one type of data, it is necessary to convert the trace data from the processor in the traces pipeline and send it to the metrics pipeline. Historically, some processors transmitted data by making use of a work-around that follows a bad practice where a processor directly exports data after processing. The connector component solves the need for this work-around and the processors that used the work around have been deprecated. On the same line, above mentioned processors are also now deprecated in recent releases and are replaced by the connectors.

Additional details about the connector’s full capabilities can be found at the following links: What are Connectors in OpenTelemetry?, OpenTelemetry Connector Configurations

The Old Architecture:

Before picture of how processors emitted data directly to another pipelines exporter

New Architecture Using a Connector:

How the pipeline should work using the connector component

Building Example Connector

For this tutorial, we will write an example connector that takes traces and converts them into metrics as a basic example of how the connector component in OpenTelemetry functions. The functionality of the basic connector is to simply count the number of spans in traces that contain a specific attribute name. The count of these occurrences are stored in the connector.

Configurations

Setting up Collector Config:

Setup the configuration you will use for the OpenTelemetry Collector in the config.yaml file. This file defines how your data will be routed, processed and exported. The configurations defined in the file, detail how you want your data pipeline to behave. You can define the components and how the data moves through your defined pipeline from start to end. There are further details about how to configure a collector at Collector Configurations.

Use the following code for the example connector we will build. The code is an example of a basic valid OpenTelemetry Collector configuration file.

receivers:
  otlp:
    protocols:
      grpc:
        endpoint: 0.0.0.0:4317
      http:
        endpoint: 0.0.0.0:4318

exporters:
  # NOTE: Prior to v0.86.0 use the `logging` instead of `debug`.
  debug:

connectors:
  example:

service:
  pipelines:
    traces:
      receivers: [otlp]
      processors: [batch]
      exporters: [example]
    metrics:
      receivers: [example]
      exporters: [debug]

In the connectors portion of the above code, you need to declare the names of the usable connectors for your pipeline. Here, example is the name of the connector we will create in this tutorial.

Implementation

  1. Create a folder for your example connector. In this tutorial we will create a folder called exampleconnector.

  2. Navigate to the folder and run

    go mod init github.com/gord02/exampleconnector
    
  3. Run go mod tidy

    This will create files go.mod and go.sum.

  4. Create the following files in the folder

    • config.go - A file to define the connector’s settings
    • factory.go - A file to create instances of the connector

Create your connector settings in config.go

In order to be instantiated and participate in pipelines, the collector needs to identify your connector and properly load its settings from within its configuration file.

In order to be able to give your connector access to its settings, create a Config struct. The struct must have an exported field for each of the connector’s settings. The parameter fields added will be accessible from the config.yaml file. Their name in the configuration file is set through a struct tag. Create struct and add parameters. You can optionally add a validator function to check if the given default values are valid for an instance of your connector.

package exampleconnector

import "fmt"

type Config struct {
    AttributeName string `mapstructure:"attribute_name"`
}

func (c *Config) Validate() error {
    if c.AttributeName == "" {
        return fmt.Errorf("attribute_name must not be empty")
    }
    return nil
}

Further details about mapstructure can be found at Go mapstructure.

Implement the Factory

To instantiate the object, you will need to use the NewFactory function associated with each of the components. We will use the connector.NewFactory function. The connector.NewFactory function instantiates and returns a connector.Factory and it requires the following parameters:

  • component.Type: a unique string identifier for your connector across all collector’s components of the same type. This string also acts as the name to refer to the connector by.
  • component.CreateDefaultConfigFunc: a reference to a function that returns the default component.Config instance for your connector.
  • ...FactoryOption: the slice of connector.FactoryOptions will determine what type of signal your connector is capable of processing.
  1. Create factory.go file and define the unique string to identify your connector as a global constant.

    const (
        defaultVal = "request.n"
        // this is the name used to refer to the connector in the config.yaml
        typeStr = "example"
    )
    
  2. Create the default configuration function. This is how you choose to initialize your connector object with default values.

    func createDefaultConfig() component.Config {
        return &Config{
            AttributeName: defaultVal,
        }
    }
    
  3. Define the type of the connector you will work with. This will be passed as a factory option. A connector can connect pipelines of different or similar types. We have to define the type of the exported end of the connector and the receiver end of the connector. A connector that exports traces and receives metrics is just one distinct configuration of the connector component and the order of how it is defined matters. A connector that exporters traces and receives metrics is not the same as a connector that could export metrics and receive traces.

    // createTracesToMetricsConnector defines the consumer type of the connector
    // We want to consume traces and export metrics, therefore, define nextConsumer as metrics, since consumer is the next component in the pipeline
    func createTracesToMetricsConnector(ctx context.Context, params connector.CreateSettings, cfg component.Config, nextConsumer consumer.Metrics) (connector.Traces, error) {
        c, err := newConnector(params.Logger, cfg)
        if err != nil {
            return nil, err
        }
        c.metricsConsumer = nextConsumer
        return c, nil
    }
    

    createTracesToMetricsConnector is a function that further initializes the connector component by defining its consumer component, or the next component to ingest the data after the connector transmits the data. It should be noted that the connector is not restricted to one ordered combination of types like we have here. For example, the count connector defines several of these functions for traces to metrics, logs to metrics and metrics to metrics.

    Parameters for the createTracesToMetricsConnector: {.h4}

    • context.Context: the reference to the collector’s context.Context so your trace receiver can properly manage its execution context.
    • connector.CreateSettings: the reference to some of the collector’s settings under which your receiver is created.
    • component.Config: the reference for the receiver config settings passed by the collector to the factory so it can properly read its settings from the collector config.
    • consumer.Metrics: the reference to the next consumer type in the pipeline, which is where received traces will go. This can be a processor, exporter or another connector.
  4. Write a NewFactory function that instantiates your custom factory for your connector(component).

    // NewFactory creates a factory for example connector.
    func NewFactory() connector.Factory {
        // OpenTelemetry connector factory to make a factory for connectors
        return connector.NewFactory(
        typeStr,
        createDefaultConfig,
        connector.WithTracesToMetrics(createTracesToMetricsConnector, component.StabilityLevelAlpha))
    }
    

    It should be noted that connectors can support multiple ordered combinations of data types.

Once finished, here is factory.go:

package exampleconnector

import (
    "context"

    "go.opentelemetry.io/collector/component"
    "go.opentelemetry.io/collector/connector"
    "go.opentelemetry.io/collector/consumer"
)

const (
    defaultVal = "request.n"
    // this is the name used to refer to the connector in the config.yaml
    typeStr = "example"
)


// NewFactory creates a factory for example connector.
func NewFactory() connector.Factory {
    // OpenTelemetry connector factory to make a factory for connectors

    return connector.NewFactory(
    typeStr,
    createDefaultConfig,
    connector.WithTracesToMetrics(createTracesToMetricsConnector, component.StabilityLevelAlpha))
}


func createDefaultConfig() component.Config {
    return &Config{
        AttributeName: defaultVal,
    }
}


// createTracesToMetricsConnector defines the consumer type of the connector
// We want to consume traces and export metrics, therefore, define nextConsumer as metrics, since consumer is the next component in the pipeline
func createTracesToMetricsConnector(ctx context.Context, params connector.CreateSettings, cfg component.Config, nextConsumer consumer.Metrics) (connector.Traces, error) {
    c, err := newConnector(params.Logger, cfg)
    if err != nil {
        return nil, err
    }
    c.metricsConsumer = nextConsumer
    return c, nil
}

Implementing the Trace Connector

Implement the methods from the interface component specific to the type of the component in the connector.go file. In this tutorial we will implement the Traces connector and therefore must implement the interfaces: baseConsumer, Traces and component.Component.

  1. Define the connector struct with the desired parameters for your connector

    // schema for connector
    type connectorImp struct {
        config Config
        metricsConsumer consumer.Metrics
        logger *zap.Logger
    }
    
  2. Define the newConnector function to create a connector

    // newConnector is a function to create a new connector
    func newConnector(logger *zap.Logger, config component.Config) (*connectorImp, error) {
        logger.Info("Building exampleconnector connector")
        cfg := config.(*Config)
    
        return &connectorImp{
            config: *cfg,
            logger: logger,
        }, nil
    }
    

    The newConnector function is a factory function to create an instance of a connector.

  3. Implement Capabilities method to properly implement the interface

    // Capabilities implements the consumer interface.
    func (c *connectorImp) Capabilities() consumer.Capabilities {
        return consumer.Capabilities{MutatesData: false}
    }
    

    Implement the Capabilities method to ensure your connector is of type consumer. This method defines the capabilities of the component, whether the component can mutate data or not. If MutatesData is set to true, it indicates that the connector mutates the data structures it is handed.

  4. Implement Consumer method to consume telemetry data

    // ConsumeTraces method is called for each instance of a trace sent to the connector
    func (c *connectorImp) ConsumeTraces(ctx context.Context, td ptrace.Traces) error{
    // loop through the levels of spans of the one trace consumed
        for i := 0; i < td.ResourceSpans().Len(); i++ {
            resourceSpan := td.ResourceSpans().At(i)
    
            for j := 0; j < resourceSpan.ScopeSpans().Len(); j++ {
                scopeSpan := resourceSpan.ScopeSpans().At(j)
    
                for k := 0; k < scopeSpan.Spans().Len(); k++ {
                    span := scopeSpan.Spans().At(k)
                    attrs := span.Attributes()
                    mapping := attrs.AsRaw()
                    for key, _ := range mapping {
                        if key == c.config.AttributeName {
                            // create metric only if span of trace had the specific attribute
                            metrics := pmetric.NewMetrics()
                            return c.metricsConsumer.ConsumeMetrics(ctx, metrics)
                        }
                    }
                }
            }
        }
        return nil
    }
    
  5. Optional: Implement Start and Shutdown methods to properly implement the interface only if a specific implementation is required. Otherwise, it is enough to include component.StartFunc and component.ShutdownFunc as part of the defined connector struct.

The complete connector file:

package exampleconnector

import (
    "context"
    "fmt"

    "go.uber.org/zap"

    "go.opentelemetry.io/collector/component"
    "go.opentelemetry.io/collector/consumer"
    "go.opentelemetry.io/collector/pdata/pmetric"
    "go.opentelemetry.io/collector/pdata/ptrace"
)


// schema for connector
type connectorImp struct {
    config Config
    metricsConsumer consumer.Metrics
    logger *zap.Logger
    // Include these parameters if a specific implementation for the Start and Shutdown function are not needed
    component.StartFunc
	component.ShutdownFunc
}

// newConnector is a function to create a new connector
func newConnector(logger *zap.Logger, config component.Config) (*connectorImp, error) {
    logger.Info("Building exampleconnector connector")
    cfg := config.(*Config)

    return &connectorImp{
    config: *cfg,
    logger: logger,
    }, nil
}


// Capabilities implements the consumer interface.
func (c *connectorImp) Capabilities() consumer.Capabilities {
    return consumer.Capabilities{MutatesData: false}
}

// ConsumeTraces method is called for each instance of a trace sent to the connector
func (c *connectorImp) ConsumeTraces(ctx context.Context, td ptrace.Traces) error {
    // loop through the levels of spans of the one trace consumed
    for i := 0; i < td.ResourceSpans().Len(); i++ {
        resourceSpan := td.ResourceSpans().At(i)

        for j := 0; j < resourceSpan.ScopeSpans().Len(); j++ {
            scopeSpan := resourceSpan.ScopeSpans().At(j)

            for k := 0; k < scopeSpan.Spans().Len(); k++ {
                span := scopeSpan.Spans().At(k)
                attrs := span.Attributes()
                mapping := attrs.AsRaw()
                for key, _ := range mapping {
                    if key == c.config.AttributeName {
                        // create metric only if span of trace had the specific attribute
                        metrics := pmetric.NewMetrics()
                        return c.metricsConsumer.ConsumeMetrics(ctx, metrics)
                    }
                }
            }
        }
    }
    return nil
}

Using the Component

Summary of Using OpenTelemetry Collector Builder:

You can use the OpenTelemetry Collector Builder to build your code and run it. The collector builder is a tool that enables you to build your own OpenTelemetry Collector binary. You can add or remove components (receivers, processors, connectors and exporters) to suit your needs.

  1. Follow the OpenTelemetry Collector Builder installation instructions.

  2. Write a Configuration File:

    Once installed, the next step is to create a configuration file builder-config.yaml. This file defines the collector components you want to include in your custom binary.

    Here is an example of the configuration file you can use featuring your new connector component:

    dist:
        name: otelcol-dev-bin
        description: Basic OpenTelemetry collector distribution for Developers
        output_path: ./otelcol-dev
        otelcol_version: 0.86.0
    
    
    exporters:
        - gomod:
        # Note: Prior to v0.86.0 use the `loggingexporter` instead of `debugexporter`.
        go.opentelemetry.io/collector/exporter/debugexporter v0.86.0
    
    
    processors:
        - gomod:
        go.opentelemetry.io/collector/processor/batchprocessor v0.86.0
    
    
    receivers:
        - gomod:
    go.opentelemetry.io/collector/receiver/otlpreceiver v0.86.0
    
    
    connectors:
        - gomod: github.com/gord02/exampleconnector v0.86.0
    
    
    replaces:
    # a list of "replaces" directives that will be part of the resulting go.mod
    
    # This replace statement is necessary since the newly added component is not found/published to GitHub yet. Replace references to GitHub path with the local path
    - github.com/gord02/exampleconnector => [PATH-TO-COMPONENT-CODE]/exampleconnector
    

    It is necessary to include a replace statement. The replace section since your newly created component is not published to GitHub yet. The references to the GitHub path for your component will need to be replaced with the local path to your code.

    There are further details on replacement in go at Go mod file Replace.

  3. Build Your collector binary:

    Run the builder while passing in the builder config file detailing the included connector component which will then build the custom collector binary:

    builder --config [PATH-TO-CONFIG]/builder-config.yaml
    

    This will generate the collector binary in the specified output path directory that was in your config file.

  4. Run Your collector binary:

    Now you can run your custom collector binary:

    ./[OUTPUT_PATH]/[NAME-OF-DIST] --config [PATH-TO-CONFIG]/config.yaml
    

    The output path name and name of dist is detailed in the build-config.yaml.

Additional resources on the OpenTelemetry Collector Builder: