Blogs

AWS Lambda Python Runtime — Notes

Overview

Lambda Python runtimes include Boto3 and Botocore in the execution environment. You can import them without adding them to a deployment package for simple functions.

ApproachUse when
Handler only (.zip)No third-party dependencies beyond the runtime SDK
pip install --targetBundle dependencies into a .zip deployment package
Virtual environmentIsolate dependencies locally before zipping site-packages
Lambda layerShare dependencies across multiple functions

AWS recommends packaging all dependencies (including Boto3) when you need predictable versions. Mixing a custom urllib3 (or other transitive dependency) with the runtime SDK can cause version misalignment. See Runtime-included SDK versions.

Example handler

Writes order details to S3 using an environment variable for the bucket name.

import json
import os
import logging
import boto3

logger = logging.getLogger()
logger.setLevel("INFO")

s3_client = boto3.client("s3")
bucket_name = os.environ.get("CONTENT_BUCKET")


def upload_to_s3(key, content):
    try:
        s3_client.put_object(
            Bucket=bucket_name,
            Key=key,
            Body=content,
        )
    except Exception as e:
        logger.error(f"Failed to upload content: {e}")
        raise


def lambda_handler(event, context):
    try:
        if isinstance(event, str):
            event = json.loads(event)

        order_id = event.get("order_id")
        amount = event.get("amount")

        if not bucket_name:
            raise ValueError("Missing required bucket")

        content = (
            f"OrderId: {order_id}\n"
            f"Amount: {amount}\n"
        )

        key = f"order/{order_id}.txt"
        upload_to_s3(key, content)

        logger.info(f"Saved to S3 {key} with OrderId: {order_id}")

        return {
            "statusCode": 200,
            "message": "Success",
        }
    except Exception as e:
        logger.error(f"Error processing order: {e}")
        raise

Corrections from draft: use isinstance(event, str) (not typeof), and log {e} directly in f-strings.

Deployment package with pip --target

Install dependencies into a package/ folder, zip them at the archive root, then add your handler .py file.

Bash

pip install --target ./package boto3

cd package
zip -r ../my_deployment_package.zip .
cd ..

zip my_deployment_package.zip lambda_function.py

PowerShell

pip install --target .\package boto3

Push-Location package
Compress-Archive -Path * -DestinationPath ..\my_deployment_package.zip
Pop-Location

Compress-Archive -Path lambda_function.py -Update -DestinationPath .\my_deployment_package.zip

lambda_function.py must sit at the root of the .zip alongside installed packages.

Deployment package with a virtual environment

Bash

cd my_function
python3 -m venv my_virtual_env
source ./my_virtual_env/bin/activate

pip install boto3
pip show <package_name>   # optional: confirm install location

deactivate

cd my_virtual_env/lib/python3.x/site-packages
zip -r ../../../../my_deployment_package.zip .

cd ../../../../
zip my_deployment_package.zip lambda_function.py

Replace python3.x with your venv Python version (for example python3.14).

PowerShell

cd my_function
python -m venv my_virtual_env
.\my_virtual_env\Scripts\Activate.ps1

pip install boto3
pip show <package_name>

deactivate

cd my_virtual_env\Lib\site-packages
Compress-Archive -Path * -DestinationPath ..\..\..\..\my_deployment_package.zip

cd ..\..\..\..
Compress-Archive -Path lambda_function.py -Update -DestinationPath .\my_deployment_package.zip

On Windows, venv packages live under Lib\site-packages (not lib/python3.x/site-packages).

Lambda layers

Use a layer when multiple functions share the same dependencies.

Install dependencies into python/

pip install requests -t python/

For packages with native (C/C++) extensions (for example NumPy, Pandas), target the Lambda Linux platform:

pip install numpy \
  --platform manylinux2014_x86_64 \
  --only-binary=:all: \
  -t python/

Use manylinux2014_aarch64 for arm64 functions. For full pip flags, see Working with .zip file archives for Python.

Create the layer archive

Bash:

zip -r layer.zip python/

PowerShell:

Compress-Archive -Path .\python -DestinationPath .\layer.zip

Required directory layout

python/                 # required top-level folder
├── requests/
├── boto3/
├── numpy/
└── (transitive dependencies)

If you build from a venv, paths may look like python/lib/python3.x/site-packages/. Lambda accepts any layout as long as python/ is at the root of the .zip.

Notes

__pycache__ folders

Do not include __pycache__ directories in deployment packages or layers. AWS documents this as a packaging best practice — see Using __pycache__ folders.

Pure Python vs native wheels

  • Pure Python packages: pip install --target ./package <name> on your build machine is usually enough.
  • Native extensions: use --platform, --implementation cp, --python-version, and --only-binary=:all: so wheels match the Lambda execution environment.

References

Related Posts