Decorate the Python function - Traces

Recap

We have pretty good visibility of what happens with our function from "outside". In order to understand the performance and actual work progress inside the function, we installed AWS Lambda Powertool layer which allows us to do exactly this. Now it is the time to start implementing it in the Function code.

Python code

For my example I wrote very, and I mean it, very simple code. I use it also for CI/CD lab, where I do some unit tests in the pipeline (to show the functionality), therefore the code consist a few functions inside.

So, here is the Lambda function code.

""" Demo lambda """

import json

def sumnumbers(a, b):
    """ simple function """
    result = a + b
    return result

def mutiplynumbers(a, b):
    """ simple function """
    result = a * b
    return result

def subnumbers(a, b):
    """ simple function """
    result = a - b
    return result

def divisionnumbers(a, b):
    """ simple function """
    result = a / b
    return result


def handler(event, context):
    """ A very simple Lambda function """
    numa = 456
    numb = 234
    firstoperation = sumnumbers(numa, numb)
    secondoperation = mutiplynumbers(numa, numb)
    thirdoperation = subnumbers(numa, numb)
    fourthoperation = divisionnumbers(numa, numb)
    response = "This is sum {}, multiply {}, substraction {}, division {}".format(
        firstoperation,
        secondoperation,
        thirdoperation,
        fourthoperation
    )
    return {
        'statusCode': 200,
        'body': json.dumps(response)
    }

Decorate the handler

Tracing

Let's go to the work. For AWS Lambda powertools we can decorate each function to enable its functionalities. However, there is one function which needs to be treated differently. As all of us know, it is the handler function.

In the example there is an handler function which is triggered by Lambda. We will decorate it with Tracer first. On the end we will have all elements enabled - tracing, metrics and logging (sounds like three pillars of Observability to you? It should :) ).

First, we need to import tracer from the library (which is included as layer).

from aws_lambda_powertools import Tracer

Initialize it outside the handler.

tracer = Tracer()

And finally, decorate the handler.

@tracer.capture_lambda_handler
def handler(event, context):
    """ A very simple Lambda function """

Please remember, in order to have it operational, we need to have the previous step done (previous article in this tutorial), which means proper setup in the SAM template:

Tracing needs to be enabled for Lambda function: Tracing: Active

And variable POWERTOOLS_SERVICE_NAME set, like in our example: POWERTOOLS_SERVICE_NAME: simpleFunctionService

Having that we are able to capture:

• Coldstart and create an annotation about it
• Any response (or exception) generated by handler. All these will be included as tracing metadata.

What is the difference between annotation and metadata? As documentation states:

• annotations are key-value records associated with traces and indexed by X-Ray. You ca work with them on X-Ray, create traces groups, etc.
• metadata are also key-value records, also associated with traces, but are not indexed by X-Ray.

Functions

Ok, we decorated the handler. What about the functions? We can decorate them as well, let's do it as an example here:

@tracer.capture_method
def sumnumbers(a, b):
    """ simple function """
    result = a + b
    return result

Now we are able to capture the functions as well. Let's deploy it and see what will happen in X-Ray.

All right, we deployed the code. Please remember, that if you run it thourgh any pipeline and you perform the unit tests, you have to have AWS Lambda Powertools installed. The simplest way to do it is to install the library using pip:

pip install aws-lambda-powertools

This is exactly what I did. Of course, we can test the whole Lambda function, but I do not care about it here.

Ok, the code is deployed, after the few executions of API, we should see something similar to this:

We can see we have the new information - number of annotations. Let's jump to the one where we have the highest number, in my case - 3.

When we switch to Raw view, we have very detailed information. Let's keep our focus on one important part - coldstart annotation. Yes, we have this knowledge and we can use it in the future :)

As final element here, let's add some custom annotations to our functions.

@tracer.capture_method
def sumnumbers(a, b):
    """ simple function """
    result = a + b
    tracer.put_annotation("operationStatus", "success")
    tracer.put_annotation("operationResult", result)
    return result

Let's go to X-Ray again after the deployment. Click any trace, then click any from our small functions and go to the annotations tab. You should see similar output like on the picture below.

Cover image by Hebi B. from Pixabay