How to manually emit exceptions

[silvicir]

Hi 👋

I saw in the README.md the emitCustomEvent method, which allows sending custom events.
Is there also a method that lets me manually emit exceptions, in the same way I can with custom events?

Thanks in advance!

[gunzip]

If you're already within an existing span context, when using OpenTelemetry, you can create an utility method:

import { SpanStatusCode, trace } from "@opentelemetry/api";

const trackException = (error: Error) => {
  const currentSpan = trace.getActiveSpan();
  if (currentSpan) {
    currentSpan.recordException(error);
    currentSpan.setStatus({
      code: SpanStatusCode.ERROR,
      message: error.message,
    });
  }
};

[silvicir]

The code in the comment above didn’t work for me. I didn’t find any exception on AI.

Moreover, I also noticed the following:

• Even though I set APPINSIGHTS_SAMPLING_PERCENTAGE to 100, I don’t always see all the requests and custom events in AI.
• I tried to trigger an error on an endpoint by causing a database timeout, and I noticed different behaviors between the @pagopa/azure-tracing package and the old packages (as in https://github.com/pagopa/io-wallet/tree/master/apps/io-wallet-user-func/src/infra/azure/appinsights). Below are the screenshots of what I observed on AI in Azure.

@pagopa/azure-tracing:

old configuration:

You can run some tests using: pagopa/io-wallet#514

[gunzip]

It is possible to send an exception event using this code:

import { logs, SeverityNumber } from "@opentelemetry/api-logs";
import {
  ATTR_EXCEPTION_MESSAGE,
  ATTR_EXCEPTION_STACKTRACE,
  ATTR_EXCEPTION_TYPE,
} from "@opentelemetry/semantic-conventions";

const trackException =
  (error: Error) =>
  (loggerName = "ApplicationInsightsLogger") => {
    logs.getLogger(loggerName).emit({
      attributes: {
        [ATTR_EXCEPTION_MESSAGE]: error.message,
        [ATTR_EXCEPTION_STACKTRACE]: error.stack,
        [ATTR_EXCEPTION_TYPE]: error.name,
      },
      severityNumber: SeverityNumber.ERROR,
      severityText: "ERROR",
    });
  };

trackException(new Error(...))();

However, to get a meaningful and accurate stack trace you need to parse it using code like this:
https://github.com/microsoft/ApplicationInsights-node.js/blob/main/src/logs/exceptions.ts#L127

---

Regarding the screenshots you shared: the observed behavior is correct because the Cosmos DB SDK performs 240 retries (HTTP requests), and these retries are correctly logged as dependencies.

---

About

I don’t always see all the requests and custom events in AI.

Can you share some real scenarios?