Tool to streamline the build of python lambda functions.
Juniper is a packaging tool to stream and standardize the creation of a zip artifact for a set of AWS Lambda functions.
The zip artifacts generated include the source code of the dependencies defined in a given requirements.txt file as well as any shared libraries the function depends on. With the generated artifact, a developer can deploy a lambda function either manually, through the awscli or using a cloudformation/sam template.
With Python==3.6 and Docker installed, install juniper:
> pip install juniper
In order to package your lambda functions with juniper, you need to create a manifest file.
functions: # Name the zip file you want juni to create router: # The dependencies of the router function. requirements: ./src/requirements.txt. # Include this file in the generated zip artifact. include: - ./src/lambda_function.py
The folder structure this manifest refers to looks like:
. ├── manifest.yml ├── src │ ├── requirements.txt │ ├── lambda_function.py
> juni build
Juniper creates the following artifact ./dist/router.zip 🎉
For a more comprehensive example, please take a look at our tutorial.
The juni build command will generate the lambda artifact for all the functions and layers defined in the manifest file. However, during the development process, it may be desired to only build the lambda functions that a developer is actively working on.
To build only a subset of the resources defined in the manifest use the following command:
> juni build -f <target_fn_name>
This command will build all the functions that partially match the given target_fn_name. When using a naming convention a developer has the ability to build a subset of the lambdas defined in the manifest.
Python3.7 and Beyond
By default juniper uses docker containers to package your lambda functions. Behind the scenes, juniper creates a docker-compose file from your manifest. This file is used by the build command to spawn a build container per function definition.
Since the AWS Lambda service supports multiple python runtimes, it makes sense for juniper to give you the ability to specify a docker image. With the following manifest file, you can package the router lambda using a python3.7 image.
functions: router: # Use this docker image image: lambci/lambda:build-python3.7 requirements: ./src/router/requirements.txt # Include these local modules in the artifact include: - ./src/commonlib/mylib - ./src/router_function/router
Keep in mind that not every single docker image works, for more information on the type of images supported read juniper and docker.
AWS Lambda layers is a recent service that gives a developer the ability to pre-package a set of dependencies. A lambda function can be built on top of multiple layers, either packaged by the developer, by AWS or by a third party.
To build a layer, the juniper manifest uses a new block:
layers: base: requirements: ./src/requirements/base.txt pg: requirements: ./src/requirements/postgres.txt
With this manifest, running juni build creates two layer artifacts: one with the name base and another one named pg. Lambda layers are packaged along the lambda functions defined in the manifest and the zip files are stored in the artifacts directory.
The generated artifact includes the dependencies defined in the requirements file of the lambda layer.
Each individual section supports the definition of a custom docker image. With this feature, a layer can be built using python3.7 and another one can be built using the default python interpreter; python3.6.
layers: base: image: lambci/lambda:build-python3.7 requirements: ./src/requirements/base.txt
Juniper builds the artifact for you, you can either use the layers aws cli to upload it to AWS or you can use a SAM template definition. While using a SAM template, make sure you use the AWS::Serverless::LayerVersion resource.
To see an example on how to package lambda functions with layers, juniper includes an example in the codebase called ridge.
To update the default configuration of juniper, can use the the global section of the manifest. A sample configuration looks like:
global: image: lambci/lambda:build-python3.7 output: ./build functions: router: requirements: ./src/router/requirements.txt include: - ./src/router_function/router/lambda_function.py
Setting a docker image at a global level tells juniper to package every lambda function using that image. In this example, the zip artifacts will be stored in the ./build folder instead of the ./dist; which is the default.
Using the lambci build images to create the zip artifacts for a given set of lambda functions is sufficient for most use cases. However, there are times when the base container does not have all the build libraries necessary to install a python package. In this cases running juni build fails while trying to pip install the dependencies of the function. In addition, once the libraries are installed in the container some packages require a set of binaries to work properly at runtime.
The recommended procedure to install OS libraries and include missing dependencies is to use a dockerfile to build a local docker image. The strategy is illustrated as follows:
- Create a dockerfile using one of the lambci images as a starting point
- Build a local docker image from the docker file
- Use the local image in the juniper manifest
With this startegy, the juniper manifest will look like this:
functions: router: image: custom/local_docker_image requirements: ./src/router/requirements.txt include: - ./src/router_function/router/lambda_function.py
In this case, a developer needs to build the docker image before executing the juni build command.
At this point, the developer can push the docker image to the docker hub and use the hosted version instead of the local one. This strategy separates the build of a custom image from the build of the artifacts.
If you need binaries in the final artifact, you can place these files either in the /var/task/lambda_lib/ or the /var/task/lambda_bin/ depending on your use case. Files added to the bin folder are included in the PATH, files added to the lib, are included in the LD_LIBRARY_PATH. For more information view aws layer config.
Juniper is in charge of putting the files in the lambda_bin and lambda_lib in the right place when building an artifact.
A concrete example of the configuration is outlined in the advanced section of our documentation.
To set any pip configuration parameters, create a pip.conf file and add the path to the manifest. The pipconf setting is only available at a global level and it will apply to the packaging of all the functions defined in the manifest.
global: pipconf: ./pip.conf functions: sample: requirements: ./requirements.txt include: - ./lambda_function.py
A sample pip.conf file can be seen bellow, to see the entire list of parameters visit the official pip documentation.
[global] timeout = 5 index-url = https://download.zope.org/ppix
This list defines the entire scope of Juniper. Nothing more, nothing else.
- Minimal manifest file to define packaging
- Using docker containers as a way to install dependencies and generate the artifacts
- Ability to tailor the requirements.txt per lambda
- Create an individual zip artifact for multiple lambda functions
- Ability to include shared dependencies (python modules relative to the function being packaged)
- Specify docker image to package lamdba functions using different python runtimes
- Define pip command line arguments using a pip.conf file
- Packaging of lambda layers
For guidance on setting up a development environment and how to make a contribution to Juniper, see the contributing guidelines.
Release history Release notifications
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size juniper-0.5.3-py3-none-any.whl (24.5 kB)||File type Wheel||Python version py3||Upload date||Hashes View hashes|
|Filename, size juniper-0.5.3.tar.gz (25.2 kB)||File type Source||Python version None||Upload date||Hashes View hashes|