Skip to content

Latest commit

 

History

History
380 lines (296 loc) · 17.9 KB

README.md

File metadata and controls

380 lines (296 loc) · 17.9 KB
Raw

Sample Python Monorepo [A work in progress]

This is a sample monorepo that can be used as a starting point for a serverless python project on AWS, using SAM and lambda containers.

It includes two sample lambda services:

  1. hello-world - a simple lambda handler without API, packaged as Zip (uploaded to s3 bucket)
  2. greeting - a simple lambda service with API, using Flask service framework, deployed as container (image uploaded to ECR)

The repo is great for ramping up on the following concepts:

  • Developing inside Devcontainer (dockerized development environment)
  • Using Poetry for dependency management and packaging
  • Unit testing lambda services
  • Working with Localstack
  • Building and running serverlass apps locally using sam cli
  • pre-commit hooks
  • Terraform
  • Github workflows
  • Changeset for package versions

Next Steps

The monorepo will be extendedt to support:

  • When running in AWS mode, Start Lambda with AppConfigConfigurationProvider when runs on AWS + Define IAM role to access AppConfig and SecretsManager
  • Logging using [aws-powertools] (https://github.com/aws-powertools/powertools-lambda-python)
  • Add Config Schema validators under each /config folder and deploy it in terraform
  • Use Lambda Extensions to better handle calls to AppConfig and SSM
  • Semantic release using changeset - blogpost here
  • Terraform to deploy infra on localstack and on AWS
  • Remove IAM Role and Policy resources from SAM templates
  • Add a deployment.yml workflow to deploy to AWS and to Localstack
  • Add e2e.yml workflow with a simple e2e test which runs nightly (every night)

Issues

  • Semantic released is not used until sample key concepts are implemented.

Setup Development Environment

Intro to Dev containers

In order to ease the setup time of new development environment, we are using dev containers while developing, thus by leveraging docker, we have a unified docker image that automatically installs anything that we need to be able to develop.

This requires all of our developers to use docker and Visual Studio Code, to get optimized development environment.

Prerequisites

  1. Install Git
  2. Install Visual Studio Code and install the Dev Containers Extension
  3. Install Docker for Mac or Docker for Windows and make sure the docker daemon is started

Initial Setup

  1. Clone the DevContainer repo.
  2. Open the repo in visual studio code
  3. Create and edit .devcontainer/.env based on the .env.sample file. Set TARGETARCH and VARIANT based on your computer architecture, as explained in the comments on .env.sample.
  4. Press CMD SHIFT P and then type reopen in container
  5. Once the container is ready you should have a working dev-environment.
  6. If you are running into Remote-Containers CLI: RPC pipe not configured error, please follow this fix

Dependency Management, Packaging and Versioning

Poetry is used for dependency management in the monorepo instead of pip. Potery settings and list of dependencies is managed in pyproject.toml file for each package, in addition to the root. To install a package, add <package-name> = "<version>" to the pyproject.toml file and install using poetry install. Make sure you distinguish dev and prod dependencies.

In the long run, we aim to release packages into a private Github package registry and install them from the registry. In the short term, we'll keep it simple and won't publish the packages, but rather install them locally. When develop = true, the package is installed in editable mode, so changes in the local package will be reflected directly.

For example:

[tool.poetry.dependencies]
my_local_package = { path = "../packages/my_local_package", develop = false }

Package versioning

We will use python-semantic-release to automatically manage package versioning per package. The configuration of python-semantic-release this will be centralized in the root's pyproject.toml file.

There are several approaches to bumping versions:

  1. Manual Bumping - Manually bumping version in the pyproject.toml file
  2. Selective Release Rules - Use python-semantic-release configuration options for selective release rules.
  3. Custom Release Script - For more control, a custom release script using the semantic_release library.

We will start with option 2 This approach will help starting with a basic release_rules configuration. As the project grows, we will refine the rules to handle more complex scenarios, and consider using a custom script for intricate versioning logic if needed.

Following the Semantic Versioning Convention this is how version is incremented: Given a version number MAJOR.MINOR.PATCH, increment the:

  • MAJOR version when you make incompatible API changes
  • MINOR version when you add functionality in a backward compatible manner
  • PATCH version when you make backward compatible bug fixes

Commit Messages Format (to be enforced through Code Review):

To allow automatic versioning, we'll need to follow commit message format.

  1. Commit Message Format: <type>(<scope>): <short summary>.

  • type: This denotes the kind of change the commit introduces. Common types include feat (new feature), fix (bug fix), docs (documentation changes), style (style changes that do not affect the meaning of the code), refactor (code change that neither fixes a bug nor adds a feature), test (adding missing tests or correcting existing ones), chore (updates to the build process, auxiliary tools, and libraries such as documentation generation).

  • scope (optional): A scope may be provided to indicate a more specific part of the codebase the change affects.

  • short summary: A concise description of the changes.

  • Examples: feat(authentication): add jwt support fix(database): resolve connection leak

  1. Automated Version Management python-semantic-release will analyze these commit messages from the main branch (or whichever branch is configured) to automatically determine the type of version bump required for the next release: Commits with feat will trigger a minor version bump. Commits with fix will trigger a patch version bump. Commits that include a BREAKING CHANGE footer will trigger a major version bump, regardless of the commit type. Custom rules in the [tool.semantic_release.release_rules] section can further refine how other types of commits (like docs, style, or chore) influence the version bump.

  2. Release Process Upon merging changes into the release branch (e.g., main), the CI pipeline should include a step that runs python-semantic-release. This tool will:

  • Analyze commits since the last release.
  • Determine the next version number based on the commit messages.
  • Update the version in pyproject.toml and any other configured version_variable files.
  • Generate or update the CHANGELOG.md.
  • Tag the release in the VCS (Version Control System, e.g., Git).
  • Optionally, upload the package to PyPI or another package repository if configured.

Local Dev environment

We're using VS Code DevContainer to run our dockerized development environment. To run the development environment locally, click Shift+Command+p -> Reopen in Container

We're leveraging AWS Localstack to emulate a cloud environment locally. To do so, we are using docker-compose to setup the DevContainer and Localstack containers on your host machine.

After the DevContainer is created, we run the script ./devcontainer/post_create.sh in which we install all the dependencies and build the packages. In addition, it creates two aliases, which allows running aws and sam commands against the local stack. So instead of aws or sam, you can use aws-localstack and sam-localstack.

NOTE: In order for these aliases to be available, open a new bash terminal and try them. If for some reason the aliases are not recognized, you can set them up naually as follows:

echo "alias aws-localstack='AWS_ACCESS_KEY_ID=test AWS_SECRET_ACCESS_KEY=test AWS_DEFAULT_REGION=us-east-1 AWS_ENDPOINT_URL=$CLOUD_ENDPOINT_OVERRIDE aws'" >> ~/.bash_aliases
echo "alias sam-localstack='AWS_ACCESS_KEY_ID=test AWS_SECRET_ACCESS_KEY=test AWS_DEFAULT_REGION=us-east-1 AWS_ENDPOINT_URL=$CLOUD_ENDPOINT_OVERRIDE sam'" >> ~/.bash_aliases

Once we add terraform files to create the infrastructure, we'll be able to deploy the infra on our localstack.

Local Packages

To install, build and test a local package, do the following from the repo root:

cd packages/<your-package>
poetry install
poetry run pytest

If you have package sependency, such as packages.package-a dependes on package-b, poetry can install the local packages and they can be imported. See package-a and package-b examples.

NOTE: When running on VS Code, you may see that the dependency local package is not recognized by pylance. Please ignore it.

To install all packages run from the monorepo root:

python run_script.py install-all

In general, every script defined in packages/scripts pyproject.toml can be run similarly:

python run_script.py <script-name>

Running and Testing Services Locally

There are several options to run services locally:

  1. Run the flask app directly
  2. Run local lambda using sam cli
  3. Deploy the lambda to localstack using sam cli

Note: We use samconfig.toml file to configure SAM CLI. We will use --config-env command argument to specify which configuration to use. The default configuration is used for local dev environment.

1. Running the flask application directly

cd services/<my-service>
flask --app app.py run --debug

2.Running lambda locally using SAM CLI

  • Use sam local invoke for invoking lambdas without API (e.g. hello-world).
  • Use sam start-api to invoke lambdas with API (e.g. greeting).
  • Remove this line from greeting/samconfig.tomland hello-world/samconfig.toml if you are not running on Mac: container_host = 'host.docker.internal'

Building & Invoking Hello-world lambda (Zip)

cd services/hello-world
sam build
sudo sam local invoke

Building & Invoking Greeting lambda (container)

cd services/greeting
sam build
sudo sam local start-api

Open your browser and ensure the endpoints work correctly:

  • http://127.0.0.1:3000/hello
  • http://127.0.0.1:3000/hello/Dan

3. Deploying lambda to Localstack using SAM CLI

You can deploy the lambda function to localstack similarly to deploying to aws, by using aws-localstack alias.

Note: Use local dev configuration and Zip build when deploying to localstack if you don;t have Pro Loaclstack account (ECR, AppConfig and SecretsManager are pro features)

Building and Deploying Hello-world lambda to localstack

First, create an S3 bucket in local stack for the lambdas packaged using sam cli:

aws-localstack s3api create-bucket --bucket sam-lambda-code-hello-world

Then, build and deploy:

cd services/hello-world
sam build
sam-localstack deploy

Once deployed to localstack, all lambdas are available on a single endpoint and can be invoked using function name. Get the FunctionName from the result of running:

aws-localstack lambda list-functions

Invoke the lambda on localstack with the relevant FunctionName:

aws-localstack lambda invoke \
--function-name <FunctionName> \
--payload '{}' \
--cli-binary-format raw-in-base64-out \
output.txt

To view localstack logs run:

sudo docker logs localstack-main

Deploying services to AWS

Make sure you are logged in to AWS Dev account using aws sso login. If your SSO login is not configured, run:

    aws configure sso

Skip SSO session name and set the profile name to default. If you are using multiple profiles (e.g. prod and dev), add --profile-name <profile> to each command.

Note: Deploy will automatically use the template under .aws-sam/build.

Building and Dwploying Hello-world to AWS (Zip)

aws s3api create-bucket --bucket sam-lambda-code-hello-world

Then, build and deploy:

cd services/hello-world
sam build
sam deploy

Once deployed to localstack, all lambdas are available on a single endpoint and can be invoked using function name. Get the FunctionName from the result of running:

aws lambda list-functions

Invoke the lambda on localstack with the relevant FunctionName:

aws lambda invoke \
--function-name hello-world \
--payload '{}' \
--cli-binary-format raw-in-base64-out \
output.txt

Building and Deploying Greeting lambda container

Build the container (note that in samconfig.toml the --use-container is set to true, hence no need to explicitly specify it here:). Use the greeting package version as DockerTag when deployng to AWS.

cd services/greeting
sam build --config-env <prod|dev> --parameter-overrides 'DockerTag=<package-version>'

Ensure there's an ECR repository for the lambda images, or create one. Use the service name as repository name, greeting in this example.

aws ecr create-repository --repository-name greeting --query 'repository.repositoryUri' > repoUri.txt

Get the repository URI from repoUri.txt and delete repoUri.txt.

Now, deploy to localstack using sam deploy (open a new bash window): Use prod or dev as config env. For prod, det DockerTag to be the recent greeting package version.

sam deploy --config-env <prod|dev> --parameter-overrides 'DockerTage=<package-version>'

You can list the functions and see that the newly deployed lambda appears in the list:

aws lambda list-functions

To invoke the function:

aws lambda invoke \
--function-name greeting \
--payload '{"headers": {}, "path": "/hello", "httpMethod": "GET"}' \
--cli-binary-format raw-in-base64-out \
output.txt

aws lambda invoke \
--function-name greeting \
--payload '{"headers": {}, "path": "/hello/Danny", "httpMethod": "GET"}' \
--cli-binary-format raw-in-base64-out \
output.txt

You can also go to test the service via the AWS console:

  1. Lambda dashboard -> Functions -> greeting -> greeting-service -> Greeting API
  2. Select the enfpoint
  3. Use the Test tab

If needed, a Function URL can be set, but we'll skip it for now.

Note: When deployed remotely, the greeting message returns with 5 exclamation points, according to the remote config, but when running locally or on local stack, it is returned with a single exclamation point. Make sure this is indeed what you are observing when running locally or remptly.

Testing

Three main types of tests in this repo: Unit Tests

  • Tests individual software units (functions and classes) in isolation.
  • Best Practices: Write unit tests alongside code development, ensure comprehensive coverage of the unit's logic, and strive for isolated and repeatable tests. Unit tests are required in each PR, and Code Review process should ensure it. (Later on coverage tools will be added)
  • Naming: <file-name>_test.py
  • Mark: Given most of our tests are unit tests, there's no need to mark them
  • Executed:
    • Pre-push runs all impacted unit tests
    • CI: All unit tests are executed automatically and must pass, otherwise merge to main fails
    • Nightly: All tests are executed nightly (workflow will be added soon)

Integration Tests

  • Verifies how different software units interact with each other.
  • Best Practices: Focus on interfaces and data exchange between units, test with stubs or mocks for external dependencies (except for the modules which their integration is being verified), and design tests to uncover integration issues.
  • Naming: <file-name>_integration_test.py
  • Mark: @pytest.mark.integration
  • Executed:
    • Nightly on each environment: All tests are executed nightly (workflow will be added soon)

End-to-end Tests

  • Simulates real user journeys, testing how different system components work together from start to finish. It encompasses functional, non-functional, and integration testing aspects.
  • Best Practices:
    • Design tests that mimic real user workflows and critical business processes.
    • Prioritize high-impact user scenarios that deliver the most value.
    • Use realistic and well-structured test data sets that reflect real-world conditions.
    • Ensure data isolation to prevent tests from interfering with each other.
  • Naming: /e2e/_e2e_test.py
  • Mark: @pytest.mark.e2e
  • Execeuted:
    • Nightly on each environment: All tests are executed nightly (workflow will be added soon)