Feature request: Calculate Log Sampling Per Request

[jasonwadsworth]

Use case

The log sampling concept is great but not very useful in practice if you have provisioned concurrency turned on due to the decision happening at initialization.

Solution/User Experience

Per request sampling would be ideal, as opposed to per log sampling. That would give the user the ability to see the entirety of a request that was sampled rather than a mix of logs for different requests.

Alternative solutions

No response

Acknowledgment

• This feature request meets Powertools for AWS Lambda (TypeScript) Tenets
• Should this be considered in other Powertools for AWS Lambda languages? i.e. Python, Java, and .NET

Future readers

Please react with 👍 and your use case to help us understand customer demand.

[dreamorosi]

Hi @jasonwadsworth, thanks for opening this issue.

This is an intended limitation that is documented in the docs, and we designed this feature this way to avoid adding overhead for those customers who don't use the feature.

Up until now we have never gotten any demand of changing this behavior, but it's something we could consider provided that there's enough demand from multiple customers.

In the meantime, you can call the logger.refreshSampleRateCalculation() method manually at the beginning or end of your function handler to recalculate the sampling.

[dreamorosi]

Hi @jasonwadsworth, are you by chance using the Middy.js middleware or class method decorator like:

import { Logger } from '@aws-lambda-powertools/logger';
import { injectLambdaContext } from '@aws-lambda-powertools/logger/middleware';
import middy from '@middy/core';

export const handler = middy(async () => {
  // your code
}).use(injectLambdaContext(logger);

or

import { Logger } from '@aws-lambda-powertools/logger';

class Lambda {
  @logger.injectLambdaContext()
  async handler (event, context) {
    // your code
  }
}

const lambda = new Lambda();
export const handler = lambda.handler.bind(lambda);

If yes, we could add an option to both of them to allow you to configure this behavior and opt-in into recalculating the rate at every request, for example:

import { Logger } from '@aws-lambda-powertools/logger';
import { injectLambdaContext } from '@aws-lambda-powertools/logger/middleware';
import middy from '@middy/core';

export const handler = middy(async () => {
  // your code
}).use(injectLambdaContext(logger, { refreshSampleRateCalculation: true }); // same for the decorator

This would be an unobtrusive change that we could roll out easily at any time.l

[jasonwadsworth]

I'm not, but I was going to try your first suggestion for now.

[dreamorosi]

Thank you for clarifying.

As of now, unless you're using one of these two methods then the Logger has no way of knowing when a request starts/ends, so it'd be tricky to implement.