tlttnz2018 / pino-lambda Goto Github PK

A lightweight drop-in decorator for pino that takes advantage of the unique environment in AWS Lambda functions.

This wrapper reformats the default pino output so it matches the existing Cloudwatch format. The default pino configuration loses some of the built in support for request ID tracing that lambda has built into Cloudwatch insights.

It also tracks the request id, correlation ids, and xray tracing from upstream services, and can be set to debug mode by upstream services on a per-request basis.

• Capture and forward correlation IDs through different Lambda event sources
• Decorated Lambda Handlers

pino-lambda is a drop-in replacement for pino. The same configuration and setup can be used without changes.

import pino from 'pino-lambda';
const logger = pino();

async function handler(event, context) {
  // new extension added to pino to automatically track requests across all instances of pino
  logger.withRequest(event, context);

  // typical logging methods passthrough to pino
  logger.info({ data: 'Some data' }, 'A log message');
}

Cloudwatch output will now match the native console.log output, correctly preserving @requestid, @timestamp, and @message properties for use in Cloudwatch Insights and other Cloudwatch aware tools such as Datadog and Splunk.

2018-12-20T17:05:25.330Z    6fccb00e-0479-11e9-af91-d7ab5c8fe19e    INFO  A log message
{
   "awsRequestId": "6fccb00e-0479-11e9-af91-d7ab5c8fe19e",
   "x-correlation-id": "238da608-0542-11e9-8eb2-f2801f1b9fd1",
   "x-correlation-trace-id": "Root=1-5c1bcbd2-9cce3b07143efd5bea1224f2;Parent=07adc05e4e92bf13;Sampled=1",
   "level": 30,
   "message": "Some A log message",
   "data": "Some data"
}

All instances of pino-lambda will be automatically log the request id so you don't need to pass an instance of a logger to all of your functions.

// handler.js
import pino from 'pino-lambda';
const logger = pino();

import { doSomething } from './service';

async function handler(event, context) {
  logger.withRequest(event, context);

  doSomething();
}

A second instance of the pino logger in another file automatically logs the request ID captured by the logger in the above handler. This alleviates the need to pass an instance of a logger around, or pass the context.

// service.js
import pino from 'pino-lambda';
const logger = pino();

export function doSomething() {
  logger.info({ data: 'Welp' }, 'Another log message');
}

Cloudwatch Output

2018-12-20T17:05:25.330Z    6fccb00e-0479-11e9-af91-d7ab5c8fe19e    INFO  A log message
{
   "awsRequestId": "6fccb00e-0479-11e9-af91-d7ab5c8fe19e",
   "x-correlation-id": "238da608-0542-11e9-8eb2-f2801f1b9fd1",
   "x-correlation-trace-id": "Root=1-5c1bcbd2-9cce3b07143efd5bea1224f2;Parent=07adc05e4e92bf13;Sampled=1",
   "level": 30,
   "message": "Some A log message",
   "data": "Some data"
}

By default, the following event data is tracked for each log statement.

Every AWS Lambda request contains a unique request ID, context.awsRequestId. If the request originated outside of the AWS platform, the request ID will match the event.header.x-correlation-id value. However, if the request originated from within the AWS platform, the event.header.x-correlation-id will be set to the request ID of the calling service. This allows you to trace a request across the entire platform.

Amazon XRAY also has a unique tracing ID that is propagated across the requests and can be tracked as well.

When upstream services invoke a Lambda using pino-lambda they can send the x-correlation-debug header with a value of true. This will enable debug logging for that specific request. This is useful for tracing issues across the platform.

You can use the this wrapper outside of the AWS lambda function in any place you want. This is especially useful in private npm modules that will be used by your AWS Lambda function. The default logger context is a shared instance, so it inherits all properties the default is configured for, and will emit request information for all logs. This effectively allows you to track a request across its entire set of log entries.