Observability

Monitoring

Open Telemetry

OpenTelemetry is an open-source standard for distributed tracing, which allows you to collect and analyze data about the performance of your systems. MassTransit can be configured to use OpenTelemetry to instrument message handling, so that you can collect telemetry data about messages as they flow through your system.

By using OpenTelemetry with MassTransit, you can gain insights into the performance of your systems, which can help you to identify and troubleshoot issues, and to improve the overall performance of your application.

There is a good set of examples opentelemetry-dotnet how it can be used for different cases

Tracing

ASP.NET Core application

This example is using following packages:

• OpenTelemetry.Extensions.Hosting
• OpenTelemetry.Exporter.Console

var builder = WebApplication.CreateBuilder(args);

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

builder.Services.AddOpenTelemetry()
    .ConfigureResource(ConfigureResource)
    .WithTracing(b => b
        .AddSource(DiagnosticHeaders.DefaultListenerName) // MassTransit ActivitySource
        .AddConsoleExporter() // Any OTEL suportable exporter can be used here
    );

Console application

This example is using following packages:

• OpenTelemetry
• OpenTelemetry.Exporter.Console

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

Sdk.CreateTracerProviderBuilder()
    .ConfigureResource(ConfigureResource)
    .AddSource(DiagnosticHeaders.DefaultListenerName) // MassTransit ActivitySource
    .AddConsoleExporter() // Any OTEL suportable exporter can be used here
    .Build()

That's it you application will start exporting MassTransit related traces within your application

Metrics

ASP.NET Core application

This example is using following packages:

• OpenTelemetry.Extensions.Hosting
• OpenTelemetry.Exporter.Console

var builder = WebApplication.CreateBuilder(args);

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

builder.Services.AddOpenTelemetry()
    .ConfigureResource(ConfigureResource)
    .WithMetrics(b => b
        .AddMeter(InstrumentationOptions.MeterName) // MassTransit Meter
        .AddConsoleExporter() // Any OTEL suportable exporter can be used here
    );

Console application

This example is using following packages:

• OpenTelemetry
• OpenTelemetry.Exporter.Console

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

Sdk.CreateTracerProviderBuilder()
    .ConfigureResource(ConfigureResource)
    .AddMeter(InstrumentationOptions.MeterName) // MassTransit Meter
    .AddConsoleExporter() // Any OTEL suportable exporter can be used here
    .Build()

The OpenTelemetry metrics captured by MassTransit:

Counters

Gauges

Histograms

Labels

Metric names and labels can be configured with Options:

services.Configure<InstrumentationOptions>(options =>
{
    // Configure
});

Application Insights

Azure Monitor has direct integration with Open Telemetry:

• OpenTelemetry overview
• Enable Azure Monitor OpenTelemetry for .NET

ASP.NET Core application

This example is using following packages:

• OpenTelemetry.Extensions.Hosting
• Azure.Monitor.OpenTelemetry.Exporter

var builder = WebApplication.CreateBuilder(args);

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

builder.Services.AddOpenTelemetry()
    .ConfigureResource(ConfigureResource)
    .WithTracing(b => b
        .AddSource(DiagnosticHeaders.DefaultListenerName) // MassTransit ActivitySource
        .AddAzureMonitorTraceExporter(
        {
            o.ConnectionString = "<Your Connection String>";
        }))
    .WithMetrics(b => b
        .AddMeter(InstrumentationOptions.MeterName) // MassTransit Meter
        .AddAzureMonitorMetricExporter(o =>
        {
            o.ConnectionString = "<Your Connection String>";
        }));

Console application

This example is using following packages:

• OpenTelemetry
• Azure.Monitor.OpenTelemetry.Exporter

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

Sdk.CreateTracerProviderBuilder()
    .ConfigureResource(ConfigureResource)
    .AddSource(DiagnosticHeaders.DefaultListenerName) // MassTransit ActivitySource
    .AddAzureMonitorTraceExporter(
    {
        o.ConnectionString = "<Your Connection String>";
    })
    .Build();

Sdk.CreateTracerProviderBuilder()
    .ConfigureResource(ConfigureResource)
    .AddMeter(InstrumentationOptions.MeterName) // MassTransit Meter
    .AddAzureMonitorMetricExporter(o =>
    {
        o.ConnectionString = "<Your Connection String>";
    })
    .Build()

You can also refer to the sample: Sample-ApplicationInsights

Prometheus

Open Telemetry is more preferable choice of integration

Open Telemetry integration

This example is using following packages:

• OpenTelemetry.Extensions.Hosting
• OpenTelemetry.Exporter.Prometheus
• OpenTelemetry.Exporter.Prometheus.AspNetCore

void ConfigureResource(ResourceBuilder r)
{
    r.AddService("Service Name",
        serviceVersion: "Version",
        serviceInstanceId: Environment.MachineName);
}

builder.Services.AddOpenTelemetry()
    .ConfigureResource(ConfigureResource)
    .WithMetrics(b => b
        .AddMeter(InstrumentationOptions.MeterName) // MassTransit Meter
        .AddPrometheusExporter()
    );
    
var app = builder.Build();

app.UseOpenTelemetryPrometheusScrapingEndpoint(); // Map prometheus metrics endpoint

In case you want to migrate from direct integration to using Open Telemetry, and use previous metric names, just configure them through Options:

builder.Services.Configure<InstrumentationOptions>(options =>
{
    ReceiveTotal = "mt.receive.total";
    // Configure other names by using similar approach
});

Lifetime Observers

MassTransit supports several message observers allowing received, consumed, sent, and published messages to be monitored. There is a bus observer as well, so that the bus life cycle can be monitored.

Bus

To observe bus life cycle events, create a class which implements IBusObserver. To configure a bus observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation.

services.AddBusObserver<BusObserver>();

services.AddBusObserver(provider => new BusObserver());

Receive Endpoint

To configure a receive endpoint observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation.

services.AddReceiveEndpointObserver<ReceiveEndpointObserver>();

services.AddReceiveEndpointObserver(provider => new ReceiveEndpointObserver());

Pipeline Observers

Receive

To observe messages as they are received by the transport, create a class that implements the IReceiveObserver interface, and connect it to the bus as shown below.

To configure a receive observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation. When a container is not being used, the ConnectReceiveObserver bus method can be used instead.

services.AddReceiveObserver<ReceiveObserver>();

services.AddReceiveObserver(provider => new ReceiveObserver());

Consume

If the ReceiveContext isn't fascinating enough for you, perhaps the actual consumption of messages might float your boat. A consume observer implements the IConsumeObserver interface, as shown below.

To configure a consume observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation. When a container is not being used, the ConnectConsumeObserver bus method can be used instead.

services.AddConsumeObserver<ConsumeObserver>();

services.AddConsumeObserver(provider => new ConsumeObserver());

Consume Message

Okay, so it's obvious that if you've read this far you want a more specific observer, one that only is called when a specific message type is consumed. We have you covered there too, as shown below.

To connect the observer, use the ConnectConsumeMessageObserver method before starting the bus.

The ConsumeMessageObserver<T> interface may be deprecated at some point, it's sort of a legacy observer that isn't recommended.

Send

Okay, so, incoming messages are not your thing. We get it, you're all about what goes out. It's cool. It's better to send than to receive. Or is that give? Anyway, a send observer is also available.

To configure a send observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation. When a container is not being used, the ConnectSendObserver bus method can be used instead.

services.AddSendObserver<SendObserver>();

services.AddSendObserver(provider => new SendObserver());

Publish

In addition to send, publish is also observable. Because the semantics matter, absolutely. Using the MessageId to link them up as it's unique for each message. Remember that Publish and Send are two distinct operations so if you want to observe all messages that are leaving your service, you have to connect both Publish and Send observers.

To configure a public observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation. When a container is not being used, the ConnectPublishObserver bus method can be used instead.

services.AddPublishObserver<PublishObserver>();

services.AddPublishObserver(provider => new PublishObserver());

State Machine Observers

Event

To observe events consumed by a saga state machine, use an IEventObserver<T> where T is the saga instance type.

To configure an event observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation.

services.AddEventObserver<T, EventObserver<T>>();

services.AddEventObserver<T>(provider => new EventObserver<T>());

State

To observe state changes that happen in a saga state machine, use an IStateObserver<T> where T is the saga instance type.

To configure a state observer, add it to the container using one of the methods shown below. The factory method version allows customization of the observer creation.

services.AddStateObserver<T, StateObserver<T>>();

services.AddStateObserver<T>(provider => new StateObserver<T>());