OpenTelemetry Custom Instrumentation

Add custom spans, attributes, and events to business-critical code paths beyond auto-instrumentation

When to Use

• Auto-instrumentation covers HTTP and database but not your business logic
• Need to trace specific operations: payment processing, recommendation generation, report building
• Adding business context (order ID, user tier, feature flag) to traces
• Wrapping third-party SDK calls that lack OpenTelemetry support

Instructions

1. Import the API only (

   @opentelemetry/api

   ), not the SDK, in application code. The SDK is initialized in

   instrumentation.ts

   .
2. Get a tracer once per module:

   const tracer = trace.getTracer('module-name')

   .
3. Use

   tracer.startActiveSpan

   for most cases — it sets the span as active context automatically.
4. Add attributes that help debugging: IDs, counts, flags, operation types.
5. Record exceptions with

   span.recordException(error)

   before setting error status.
6. For utility functions called from many places, create a reusable tracing wrapper.

// decorators/traced.ts — reusable tracing wrapper
import { trace, SpanKind, SpanStatusCode, Span } from '@opentelemetry/api';

const tracer = trace.getTracer('app');

export function traced<T extends (...args: any[]) => Promise<any>>(
  name: string,
  fn: T,
  options?: {
    kind?: SpanKind;
    attributes?: Record<string, string | number | boolean>;
  }
): T {
  return (async (...args: any[]) => {
    return tracer.startActiveSpan(
      name,
      {
        kind: options?.kind ?? SpanKind.INTERNAL,
        attributes: options?.attributes,
      },
      async (span: Span) => {
        try {
          const result = await fn(...args);
          span.setStatus({ code: SpanStatusCode.OK });
          return result;
        } catch (error) {
          span.recordException(error as Error);
          span.setStatus({ code: SpanStatusCode.ERROR, message: (error as Error).message });
          throw error;
        } finally {
          span.end();
        }
      }
    );
  }) as T;
}

// Usage
const processPayment = traced('payment.process', async (orderId: string, amount: number) => {
  const span = trace.getActiveSpan();
  span?.setAttribute('payment.order_id', orderId);
  span?.setAttribute('payment.amount', amount);

  const result = await paymentGateway.charge(orderId, amount);

  span?.setAttribute('payment.transaction_id', result.transactionId);
  span?.addEvent('payment.charged', { 'payment.method': result.method });

  return result;
});

// services/recommendation-service.ts — inline custom spans
import { trace, SpanStatusCode } from '@opentelemetry/api';

const tracer = trace.getTracer('recommendation-service');

export async function getRecommendations(userId: string, limit: number): Promise<Product[]> {
  return tracer.startActiveSpan('recommendations.generate', async (span) => {
    span.setAttribute('user.id', userId);
    span.setAttribute('recommendations.limit', limit);

    // Step 1: Get user profile
    const profile = await tracer.startActiveSpan(
      'recommendations.loadProfile',
      async (profileSpan) => {
        const result = await userService.getProfile(userId);
        profileSpan.setAttribute('user.segment', result.segment);
        profileSpan.end();
        return result;
      }
    );

    // Step 2: Score candidates
    const scored = await tracer.startActiveSpan('recommendations.score', async (scoreSpan) => {
      const candidates = await productService.getCandidates(profile.segment);
      scoreSpan.setAttribute('candidates.count', candidates.length);
      const results = await mlService.score(candidates, profile);
      scoreSpan.addEvent('scoring.complete', { top_score: results[0]?.score ?? 0 });
      scoreSpan.end();
      return results;
    });

    span.setAttribute('recommendations.returned', Math.min(scored.length, limit));
    span.end();
    return scored.slice(0, limit);
  });
}

Details

API vs SDK separation: Application code imports

@opentelemetry/api

(zero-cost no-op if no SDK is registered). The SDK is configured once in

instrumentation.ts

. This means your business code has no dependency on the SDK — if OpenTelemetry is not initialized, spans are silently discarded.

Getting the active span (without creating a new one):

import { trace } from '@opentelemetry/api';

function enrichSpan(orderId: string) {
  const span = trace.getActiveSpan();
  if (span) {
    span.setAttribute('order.id', orderId);
  }
}

Span links: Connect related traces that are not parent-child (e.g., a batch job processing items from different requests):

const span = tracer.startSpan('batch.process', {
  links: items.map((item) => ({
    context: item.spanContext,
    attributes: { 'batch.item_id': item.id },
  })),
});

What to instrument (rule of thumb):

• Business operations (order creation, payment, fulfillment) — always
• Service boundaries (incoming requests, outgoing calls) — auto-instrumentation handles this
• Significant decision points (feature flag evaluation, A/B test assignment) — as events
• Every function call — never (too much overhead, noisy traces)

Source

https://opentelemetry.io/docs/languages/js/instrumentation/

Process

1. Read the instructions and examples in this document.
2. Apply the patterns to your implementation, adapting to your specific context.
3. Verify your implementation against the details and edge cases listed above.

Harness Integration

• Type: knowledge — this skill is a reference document, not a procedural workflow.
• No tools or state — consumed as context by other skills and agents.

Success Criteria

• The patterns described in this document are applied correctly in the implementation.
• Edge cases and anti-patterns listed in this document are avoided.