Sentry-javascript: Sentry.captureException only sends event to Sentry once (AWS Lambda)

[Nickforall]

Package + Version

@sentry/node

Version:

4.6.3

Description

I'm trying to integrate the new sentry SDK into our serverless architecture, we have a general try-catch around every handler, that calls captureException. In a nutshell this is what our code looks like (I left out our business logic and a lot of utility methods).

Sentry.init({
  dsn: process.env.sentryUrl,
  environment: process.env.NODE_ENV,
  // debugging issue where we only get an event once
  beforeSend: (event, _) => { console.log(event.exception); return event }
});

async function wrapper(...) {
   try {
     handler(...)
   } catch(error) {
      Sentry.configureScope(scope => {
        scope.setTag("test", "Hello, world")
      })

      const eventId = Sentry.captureException(error)
      // debugging, log event id to see if captureException works
      console.log(eventId)

      const flushResult = await Sentry.flush(1000)
      console.log(flushResult) // see if flush succeeded
   }
}

I made the handler always throw an error. When I do the first request, everything ends up in Sentry just fine. On the second request, nothing happens. When I look in CloudWatch, this is what I find. I have added comments for convenience.

First request:

/// the eventId from captureException
2019-02-26T13:47:22.584Z: 8cf83e38081a4672841a6f795ce8105a 
/// the middleware defined in Sentry.init
2019-02-26T13:47:23.153Z: { values: [ { stacktrace: [Object], type: 'Error', value: 'This is an error' } ] } 
/// result of Sentry.flush
2019-02-26T13:47:23.294Z: true 

Second request:

/// the eventId from captureException
2019-02-26T13:47:24.761Z fbc2c8b2a4d5447380a0f92dd87896f6
/// result of Sentry.flush
2019-02-26T13:47:24.963Z: true 

As you can see, it never hits the beforeSend middleware on the second request, so I guess it's never sent to Sentry at all.

I should note that when I do Sentry.getCurrentHub().getClient().captureException(error), it does send the exception to Sentry every time, but the scope configured in Sentry.configureScope is never applied to the event.

I have tried reproducing this locally, but it only happens when deployed on AWS Lambda.

[kamilogorek]

There's a chance that it's getting de-duplicated.
You can use debug: true in your init call to see what's going on.

If it indeed is getting deduped, you can remove this integration using:

Sentry.init({
  integrations: integrations => {
    return integrations.filter(integration => integration.name !== 'Dedupe');
  }
});

https://docs.sentry.io/platforms/node/default-integrations/

[cruisercohen]

Interesting. Found this looking for a similar issue. The remove dedupe fixed it for me.

[HazAT]

Closing this since it's not a bug.
Additionally the next major version makes Dedupe optional, which means there is no deduping by default.