Wide Logger

A canonical wide logger that is built to gather key, value pairs and then flush them all to the console in a single log message. This does not replace your existing detailed debug logging, it is an addition. All logs emitted by the Wide Logger will be prefixed by WIDE so you can quickly and easily find them or use filtered subscriptions to record these in a single place for easy searching and correlation.

Usage

Create at the start, add anywhere, flush at the end.

import { WideLogger } from '@serverless-dna/wide-logger';

const wideLogger = new WideLogger()

const serviceProcessor = async (data: Record<string, any>) => {
    
    try {
        wideLogger.add('startEpoch', Date.now());
        wideLogger.add('premiumUser', data.premiumUser ?? false);

        // ... do some processing here
        if (data.premiumUser && data.premium.goupA) {
            wideLogger.add('premiumGroup', 'A');
        }
    } finally {
        wideLogger.flush();
    }
}

The log output will be:

WIDE {"startEpoch":1709335919083,"premiumUser":true,"premiumGroup:"A"}

When the Wide Logger is flushed it will do so using console.log() and will use the LogFormatter passed with a default of JSON (Object) output.

Included are a JsonFormatter and KeyValueFormatter. You can create your own if you have a specific format requirement.

import { WideLogger } from '@serverless-dna/wide-logger';
import { KeyValueFormatter } from '@serverless-dna/wide-logger';

const wideLogger = new WideLogger(new KeyValueFormatter())

const serviceProcessor = async (data: Record<string, any>) => {
    try {
        wideLogger.add('startEpoch', Date.now());
        wideLogger.add('premiumUser', data.premiumUser ?? false);

        // ... do some processing here
        if (data.premiumUser && data.premium.goupA) {
            wideLogger.add('premiumGroup', 'A');
        }
    } finally {
        wideLogger.flush();
    }
}

Using the KeyValueFormatter will output the following:

WIDE startEpoch=1709335919083|premiumUser=true|premiumGroup="A"

The KeyValueFormatter also accepts a separator character in the constructor so you can use your own custom separator for your log ingestion engine with a default of '|'.

Middy Middleware Usage

We have also packaged a middleware class for MiddyJs users to simplify your WideLogging needs for straight lambda function calls.

import middy from '@middy/core';
import { WideLogger, WideLoggerMiddy } from '@serverless-dna/wide-logger';

const wideLogger = new WideLogger()

const lambdaHandler = (event, context) => {
    wideLogger.add('service', 'mySpecialService')
    wideLogger.add('startEpoch', Date.now());
  /* your business logic */

}

export const handler = middy()
  .use(WideLoggerMiddy(wideLogger))
  .handler(lambdaHandler)

Injecting AWS Lambda Context

You can optionally include relevant details from the AWS Lambda function context within each Wide log entry. This will be added as a single key/object value pair. The KeyValueFormatter uses JSON.stringify to format values, so if you use this formatter be aware that the lambda context is an encoded string value.

The context injected has the following structure.

{
    "lambdaContext":
    {
        "lambdaFunction":
        {
            "arn": "arn:aws:lambda:us-east-1:123456789012:function:your-lambda-function",
            "name": "your-lambda-function",
            "memoryLimitInMB": "256",
            "version": "$LATEST"
        },
        "awsAccountId": "123456789012",
        "awsRegion": "us-east-1",
        "correlationIds":
        {
            "awsRequestId": "017100a8-3d65-40bb-aae1-673367af29b3",
            "xRayTraceId": "017100a8-3d65-40bb-aae1-673367af29b3"
        },
        "remainingTimeInMillis": 2999
    }
}

You can enable injecting the AWS Lmabda Context into your Wide log messages as shown here.

import middy from '@middy/core';
import { WideLogger, WideLoggerMiddy } from '@serverless-dna/wide-logger';

const wideLogger = new WideLogger()

const lambdaHandler = (event, context) => {
    wideLogger.add('service', 'mySpecialService')
    wideLogger.add('startEpoch', Date.now());
  /* your business logic */

}

export const handler = middy()
  .use(WideLoggerMiddy(wideLogger, { injectLambdaContext: true }))
  .handler(lambdaHandler)

Why Wide Logs?

Traces, Log entries and metrics are great and form an essential part of your monitoring tool-kit for the more technical aspects of your system, for example CPU usage, Throttling, Memory usage. But there is more to observing a distributed system so you need to log meta-data about your business functions, users, external dependencies, and other key criteria for enabling you to search your logs and discover the more Unknown Unknowns.

Wide logging is a technique where you write a single log message per transaction, per service. With the WideLogger class you can add key-value metadata at any point it makes sense in your code - perhaps it's meta-data about a specific if statement that is run, the latency of an external dependency, the duration of distributed service processing or perhaps it's a version stamp linking back to the actual release artifact for your service. It can be anything at all - the more you pack in, the more cardinality you will get when trying to search data about what your system is doing. At the end of your transaction call the flush() method to write your wide log entry.