Routing logs to different destinations

Learn how to route logs from a single ILogger to different OTLP destinations using a custom processor.

Você está visualizando a versão em versão em inglês desta página porque ela ainda não foi traduzida. Possui interesse em ajudar? Veja como contribuir.

This guide shows how to route logs from a single ILogger to different OTLP endpoints using a custom processor. It follows the Routing pattern described in the OpenTelemetry specification supplementary guidelines.

Why route logs?

In some scenarios, you want all application code to use the same ILogger pipeline, yet send certain logs to one backend and the rest to another. For example:

  • Logs from payment components should go to a dedicated collector endpoint (OTLP2).
  • All other logs should go to the default endpoint (OTLP1).

If your application can instead create multiple ILoggerFactory instances and have callers pick the appropriate one, consider using a dedicated logging pipeline instead.

How it works

The routing decision is made at the processor level by inspecting the CategoryName of each LogRecord. A custom processor checks whether the category name starts with a configured prefix and forwards the record to the appropriate export pipeline:

ILogger (single pipeline)
   |
   v
LoggerProvider
   |
   v
RoutingProcessor (custom)
   +-- CategoryName starts with prefix --> ExportProcessor -> OtlpLogExporter (OTLP2)
   +-- otherwise --------------------------> ExportProcessor -> OtlpLogExporter (OTLP1)
  1. Two OtlpLogExporter instances are created, each pointing at a different endpoint.
  2. Each exporter is wrapped in a BatchLogRecordExportProcessor.
  3. A custom RoutingProcessor extends BaseProcessor<LogRecord> and overrides OnEnd. It checks if the log record’s CategoryName starts with a configured prefix to decide which inner processor receives the record.
  4. The routing processor is registered on the LoggerProvider via AddProcessor.

Implementation

The custom routing processor

The RoutingProcessor inspects each log record’s CategoryName and forwards it to one of two inner processors. It also delegates ForceFlush, Shutdown, and Dispose so that both export pipelines are properly drained and cleaned up:

using OpenTelemetry;
using OpenTelemetry.Logs;

internal sealed class RoutingProcessor : BaseProcessor<LogRecord>
{
    private readonly string categoryPrefix;
    private readonly BaseProcessor<LogRecord> defaultProcessor;
    private readonly BaseProcessor<LogRecord> paymentProcessor;

    public RoutingProcessor(
        string categoryPrefix,
        BaseProcessor<LogRecord> defaultProcessor,
        BaseProcessor<LogRecord> paymentProcessor)
    {
        this.categoryPrefix = categoryPrefix ?? throw new ArgumentNullException(nameof(categoryPrefix));
        this.defaultProcessor = defaultProcessor ?? throw new ArgumentNullException(nameof(defaultProcessor));
        this.paymentProcessor = paymentProcessor ?? throw new ArgumentNullException(nameof(paymentProcessor));
    }

    public override void OnEnd(LogRecord data)
    {
        if (data.CategoryName?.StartsWith(this.categoryPrefix, StringComparison.Ordinal) == true)
        {
            this.paymentProcessor.OnEnd(data);
        }
        else
        {
            this.defaultProcessor.OnEnd(data);
        }
    }

    protected override bool OnForceFlush(int timeoutMilliseconds)
    {
        var result1 = this.defaultProcessor.ForceFlush(timeoutMilliseconds);
        var result2 = this.paymentProcessor.ForceFlush(timeoutMilliseconds);
        return result1 && result2;
    }

    protected override bool OnShutdown(int timeoutMilliseconds)
    {
        var result1 = this.defaultProcessor.Shutdown(timeoutMilliseconds);
        var result2 = this.paymentProcessor.Shutdown(timeoutMilliseconds);
        return result1 && result2;
    }

    protected override void Dispose(bool disposing)
    {
        if (disposing)
        {
            this.defaultProcessor.Dispose();
            this.paymentProcessor.Dispose();
        }

        base.Dispose(disposing);
    }
}

Registering the routing processor on the LoggerProvider

Register the routing processor on the LoggerProvider. Both loggers below share the same ILoggerFactory and LoggerProvider pipeline, but their log records are routed to different OTLP destinations based on category name:

using Microsoft.Extensions.Logging;
using OpenTelemetry;
using OpenTelemetry.Exporter;
using OpenTelemetry.Logs;

// Create two OTLP exporters pointing at different destinations.
var otlpExporter1 = new OtlpLogExporter(new OtlpExporterOptions
{
    Endpoint = new Uri("http://localhost:4317"), // OTLP destination 1
});

var otlpExporter2 = new OtlpLogExporter(new OtlpExporterOptions
{
    Endpoint = new Uri("http://localhost:4318"), // OTLP destination 2
});

// Wrap each exporter in a BatchLogRecordExportProcessor.
var defaultExportProcessor = new BatchLogRecordExportProcessor(otlpExporter1);
var paymentExportProcessor = new BatchLogRecordExportProcessor(otlpExporter2);

// Build the routing processor. Logs whose category name starts with
// "Payment." are sent to OTLP2; everything else goes to OTLP1.
var routingProcessor = new RoutingProcessor(
    categoryPrefix: "Payment.",
    defaultProcessor: defaultExportProcessor,
    paymentProcessor: paymentExportProcessor);

var loggerFactory = LoggerFactory.Create(builder =>
{
    builder.AddOpenTelemetry(logging =>
    {
        logging.AddProcessor(routingProcessor);

        // Optional: also add a console exporter so you can see all logs locally.
        logging.AddConsoleExporter();
    });
});

// Both loggers share the same ILoggerFactory / LoggerProvider pipeline.
var orderLogger = loggerFactory.CreateLogger("Order.Processing");
var paymentLogger = loggerFactory.CreateLogger("Payment.Processing");

orderLogger.LogInformation("Processing order {OrderId}.", "ORD-001");     // --> OTLP1
paymentLogger.LogInformation("Processing payment {PaymentId}.", "PAY-001"); // --> OTLP2
orderLogger.LogInformation("Order {OrderId} completed.", "ORD-001");      // --> OTLP1

// Dispose logger factory before the application ends.
// This will flush the remaining logs and shutdown the logging pipeline.
loggerFactory.Dispose();

Key considerations

  • Routing condition is evaluated per log record. Keep the logic fast – it runs synchronously on every log emit.
  • Lifecycle management. Make sure the routing processor delegates ForceFlush, Shutdown, and Dispose to all inner processors so that every export pipeline is properly drained and cleaned up.
  • Single pipeline. Because there is a single ILoggerFactory and LoggerProvider, this approach works well with dependency injection where components are not aware of multiple loggers.

Further reading


Última modificação May 19, 2026: [dotnet] sync docs with upstream opentelemetry-dotnet (#9883) (2140e850)