Issue titles from errors captured with @sentry/aws-serverless differ from @sentry/node

[s1gr1d]

Description

Originally posted in this comment by @rowbot-weisguy

@relm923 good point. We use CDK so I've been bundling it using esbuild via aws-cdk-lib/aws-lambda-nodejs.

Today, given your advice, I set up a separate layer and marked express as an external module. It works! Warning is gone and I managed to get traces that I wanted to see from the lambda as long as I explicitly include the integrations I care about (e.g. expressIntegration, redisIntegration, and postgresIntegration).

I'm still noticing the generated issue titles (span names?) are a little unhelpful.

In development, using @sentry/node I see:

Lovely! The path is so helpful to know immediately. Whereas in deployed lambdas, using @sentry/aws-serverless I see:

So I tried implementing your suggestion @andreiborza:

// instrument.lambda.ts
import * as Sentry from "@sentry/aws-serverless";
import { nodeProfilingIntegration } from "@sentry/profiling-node";

import { env } from "./env";

Sentry.init({
  dsn: env.PILOT_API_SENTRY_DSN,
  defaultIntegrations: false,
  integrations: [
    ...Sentry.getDefaultIntegrations({}).filter(
      (integration) => !["Aws", "AwsLambda"].includes(integration.name),
    ),
    nodeProfilingIntegration(),
    Sentry.expressIntegration(),
    Sentry.redisIntegration(),
    Sentry.postgresIntegration(),
  ],
  environment: env.PILOT_API_SENTRY_ENVIRONMENT,
  tracesSampleRate: Number(env.PILOT_API_SENTRY_TRACING_SAMPLE_RATE),
  profilesSampleRate: Number(env.PILOT_API_SENTRY_PROFILING_SAMPLE_RATE),
});

export const wrapHandler = Sentry.wrapHandler;
This change made no visible difference to spans / names, and I'm not sure if there's a different way I should disable the Aws integration.

Would love to make sure our Sentry issues are informative at a glance, so please let me know if there's something else I can do here. This is my first time instrumenting Sentry w/ Express & Lambda, so I have no experience with previous versions to share.

[s1gr1d]

@rowbot-weisguy The @sentry/aws-serverless SDK sets the transaction name based on the function name we get from the AWS Lambda context (see code here).

That's the reason why the naming is different to the @sentry/node SDK which you are using in development mode. Does the naming resemble the AWS Lambda function name or is this not correct either?

[rowbot-weisguy]

@s1gr1d thanks for creating a new issue for me. That's correct, the transaction names are using our lambda's name.

Is it possible to have it use the request path instead?

[s1gr1d]

You can change the transaction name based on what is documented here: https://docs.sentry.io/platforms/javascript/guides/aws-lambda/enriching-events/transaction-name/

Does that work for your case?