Skip to content

Latest commit

 

History

History
84 lines (62 loc) · 3.96 KB

PYTHON_README.md

File metadata and controls

84 lines (62 loc) · 3.96 KB
Raw

STEP-BY-STEP GUIDE FOR PYTHON-BASED DOCKER IMAGES FOR LAMBDA FUNCTIONS

Understanding the files structure

The following tree structure explains the content of the lambdas/python folder:

.
└── python/  # Folder for Python-based Lambda solution
    ├── Dockerfile          # Dockerfile that builds the Docker image
    ├── lambda_function.py  # Main "entrypoint" for the Lambda solution
    ├── requirements.txt    # Python requirements (libraries) for the imports
    ├── sample_code_1.py    # "Business logic code" 1 (used in "lambda_function.py")
    └── sample_code_2.py    # "Business logic code" 2 (used in "lambda_function.py")

Dockerfile explained

The Dockerfile uses the official AWS Python Base Image (Container Image), which is explained in detail at aws-lambda-base-image:

FROM public.ecr.aws/lambda/python:3.9

Then, a custom label for traceability is added (optional):

LABEL maintainer="Santiago Garcia Arango [san99tiago]"

Afterwards, one of the most important steps happen. This is copying the actual code into Lambda Function folder that is dynamically used for the execution (in this case ${LAMBDA_TASK_ROOT} is an pre-defined environment variable, that based on the source image creation, it corresponds to /var/task, as can be seen at python3.9 (base image)):

# Copy function's code
COPY . ${LAMBDA_TASK_ROOT}

Subsequently, we install the necessary Python dependencies (or external libraries to import), this is usually done via the pip3 install command. The --target flag specifies where to install it, so we select again the ${LAMBDA_TASK_ROOT}, for compliance reasons with the Lambda runtime specifications.

# Install python dependencies
COPY requirements.txt .
RUN pip3 install -r requirements.txt --target "${LAMBDA_TASK_ROOT}"

Finally, we finish the Dockerfile with the CMD command that specifies which entrypoint is going to be used for running the Lambda function (the handler):

# Set the CMD to your handler (could also be done as a parameter override outside of the Dockerfile)
CMD [ "lambda_function.handler" ]

Important remarks

  1. The CDK Stack creates the Lambda-Python (Docker-based) image by indicating the source folder of the necessary files that should exist in the image creation. This usually needs at least the Dockerfile and the lambda_function.py files (bare minimum). Then, we must specify at the CDK Stack the following:

    PATH_TO_FUNCTION_FOLDER = "/path/to/folder/with/necessary/files"
self.docker_lambda_function = aws_lambda.DockerImageFunction(
    code=aws_lambda.DockerImageCode.from_image_asset(PATH_TO_FUNCTION_FOLDER),
    # <Other attributes missing>
)

  2. You must match the "filename" and "function" of the Lambda entrypoint specified at the lambdas/python/Dockerfile file, with the corresponding Python file that contains the actual code at lambdas/python/lambda_function.py:

    # In /lambdas/python/Dockerfile
CMD [ "lambda_function.handler" ]
    # In /lambdas/python/lambda_function.py
def handler(event, context):
    # <code>

    See how the lambda_function corresponds to lambda_function.py and handler corresponds to hander(event, context)

  3. Even if it sounds funny, remember to have Docker installed and running, otherwise you could get errors like (which shows a failure, but the real reason why it failed, is because it couldn't use Docker commands):

    <ERROR CODE>
fail: docker login --username AWS --password-stdin https://XXXXXXXXXXXX.dkr.ecr.us-east-1.amazonaws.com exited with error code 1