A packaging tool for Python based AWS serverless applications.
Switch branches/tags
Clone or download
brysontyrrell Merge pull request #17 from brysontyrrell/develop
Support override of upload directory in S3 bucket
Latest commit 4a832d7 Oct 1, 2018



Python AWS SAM --> PyAWSSAM --> Pawssam --> Possum


Possum can be installed from the Python Package Index:

$ pip install possum

Possum requires Python 3.6+ and pipenv (pipenv must be installed separately and is not installed with Possum).


Possum is a packaging tool for serverless Python-based applications using the AWS Serverless Application Model (AWS SAM).

The sam package or aws cloudformation package options have a limitation when it comes to Python Lambda functions: they have no means of packaging external dependencies. This means that developers are always on their own for creating those Lambda artifacts, uploading them to S3, and deploying.

Possum aims to serve as a replacement to the basic package functions. The tool is based upon my approach to serverless AWS applications (opinionated) and may not be a fit for all parties.

AWS Credentials

Possum uses the Boto3 SDK for uploading artifacts to S3. You can set your AWS access and secret keys in your environment variables as described in the Boto3 documentation. Possom also accept a profile name for your AWS credentials file via the -p/--profile argument.

$ possum '<s3-bucket-name>' --profile '<my-profile-name>'

Basic Usage

Run Possum from the repository directory containing the serverless application.

$ possum '<s3-bucket-name>'

The above command will package the Python Lambda functions and upload them to S3 assuming the template file is named template.yaml. You can specify the template's name with the -t/--template argument:

$ possum '<s3-bucket-name>' -t my-template.yml

The generated deployment template will printed on the screen.

By default, Possum will upload new artifacts to a directory in your chosen S3 bucket named possum-0123456789 where the numerical value is the current timestamp.

If you wish to override this default and specify the directory path to upload new artifact, append it to the S3 bucket name using forward slashes:

$ possum '<s3-bucket-name>/<my_path>'

You can also specify the deployment template be written to a file by passing a name to the -o/--output-template argument:

$ possum '<s3-bucket-name>' -o deployment.yaml

Possum uses hashes of your function directories to determine if changes have occurred since the last run of the command. Hashes and S3 URIs are saved in a ~/.possum directory for each project you package with Possum.

To force Possum to build all functions and skip the hash check, use the -c/--clean argument.

You can view the options and instructions for using Possum with the -h argument:

$ possum -h
usage: possum [-h] [-t template] [-o output] [-p profile_name] [-c] [--docker]
          [--docker-image image_name] [-v]

Possum is a utility to package Python-based serverless applications using the
Amazon Serverless Application model with per-function dependencies.

positional arguments:
  s3_bucket             The S3 bucket to upload artifacts. You may optionally
                        pass a path within the bucket to store the Lambda
                        artifacts (defaults to 'possum-{timestamp}').

optional arguments:
  -h, --help            show this help message and exit
  -t template, --template template
                        The filename of the SAM template.
  -o output, --output-template output
                        Optional filename for the output template.
  -p profile_name, --profile profile_name
                        Optional profile name for AWS credentials.
  -c, --clean           Build all Lambda packages, ignoring previous run.
  --docker              Build Lambda packages within a Docker container
  --docker-image image_name
                        Specify a Docker image to use (defaults to
  -v, --version         Display version information.

Docker Support

The installation of some Python packages differ based on the underlying system (cryptography.io is an example). To ensure your installed dependencies are fully compatible with the Lambda environment, you may opt to run Possum within a Docker container.

The included Dockerfile in this project will create a compatible default image to use. Run the following command from the same directory as the Dockerfile to build the image:

$ docker build . -t possum:latest

This image is based upon lambci/lambda:build-python3.6. You may build your own custom image and specify it using the --docker-image argument. If you decide to use your own image it must have pipenv and possum installed!

Launch Possum in a container using the --docker argument:

$ possum '<s3-bucket-name>' --docker

Serverless App Repository Example

Here is an example of a serverless Python application with multiple Lambda functions in a single repository:

    |   |
    |   |___function1.py

For each AWS Lambda function defined in the template, Possum references the Properties:CodeUri key for the path to the function's directory.

Possum will display a warning if the function's Properties:Runtime value does not match python*. You will need to package these remaining functions separately.

The contents of each functions' directory will be copied to a temporary build directory. If a Pipfile/Pipfile.lock or requirements.txt exist, the external packages will be installed into the build directory. The entire contents of the build directory will then be zipped into a deployable Lambda artifact.

All artifacts will be uploaded to the provided S3 bucket. The imported template will be updated with the S3 locations for each Lambda function and written stdout or a file if the -o argument was provided.

The generated deployment template can be used with sam deploy or aws cloudformation deploy to deploy the application.