How to debug errors in Lambda functions

Troubleshooting production issues in Lambda environments is often challenging. CloudWatch is not easy to use to quickly identify and resolve issues. CloudWatch requires me to go into logs, search for relevant terms that I may not always know and has hard to consume stack traces.

I am using Sentry to instrument my Lambda Functions code in order to report error stack traces and associated debugging context.

Here's how I instrumented my Node function. If you are using Python environment, there are parallel docs.

Step 1: Create a dummy function to test the error reporting - I created a simple function that calls an unknown function.

const mySyncHandler = (event, context, callback) => {

  notDefinedFunc();
};

Step 2: Add Sentry to the Node folder - I used npm for this but you can use yarn as well

npm install --save @sentry/serverless

Step 3: Initialize Sentry in my Lambda function - This required me to import Sentry's serverless package and wrap. my handler.

const Sentry = require("@sentry/serverless");

Sentry.init({
  dsn:
    "Get your dsn by creating a free account with Sentry",
});

const mySyncHandler = (event, context, callback) => {

  notDefinedFunc();
};

exports.handler = Sentry.AWSLambda.wrapHandler(mySyncHandler);

That's it. I now have visibility into errors from my lambda functions.

The stack traces are detailed and point me to the line of code that's causing the exception.

Function context like aws_request_id and remaining_time_in_milis are super helpful. In addition, I have deep links to CloudWatch logs.

In addition, I made couple customizations to my error reporting:

1. I really care about timeout issues. They end up bleeding into unknown issues that are difficult to debug. So, I enabled timeout warnings.
   

   exports.handler =     Sentry.AWSLambda.wrapHandler(yourHandler, {
     captureTimeoutWarning: false,
   });
   

2. Also, while Sentry automatically captured breadcrumbs from console logs, I added some of my own
   

Sentry.addBreadcrumb({
  category: "auth",
  message: "Authenticated user " + user.email,
  level: Sentry.Severity.Info,
});

With this instrumentation, I have visibility that's unparalleled to any other solution out there. I can quickly identify issues, reproduce issues in my local environment and get to root cause of the problem.

Please share if you have used a better way to monitor errors for your Lambda functions.