Developing locally an aws lambda

The problem (as always)

In my career, I’ve often seen AWS Lambdas deployed directly without the source code being committed to GitHub. Even worse, many projects lack a defined architecture or tooling that allows testing Lambdas locally before deployment.

Here, I’ll walk through how to introduce some basic tools to fix this.

Setup aws locally

To run and debug Lambdas locally, we need an execution environment that mimics AWS services. For this, I recommend using LocalStack with Docker:

services: localstack: image: localstack/localstack ports: - "4566:4566" - "4510-4559:4510-4559" environment: - SERVICES=lambda,logs - DEBUG=1 - LAMBDA_EXECUTOR=docker - DATA_DIR=/tmp/localstack/data - AWS_DEFAULT_REGION=eu-north-1 - LOCALSTACK_API_KEY=${LOCALSTACK_API_KEY-} - USE_SSL=false volumes: - "./localstack/volume:/var/lib/localstack" - "/var/run/docker.sock:/var/run/docker.sock" 

There are 2 things you need to keep notice:

1. Upon SERVICES environmental variable we need to define 2 services:
   • lambda that indicates we need to use localstack for lambda execution.
   • logs in which any log outpud of the lambda would be outputed into stdout. This is the equivalent of cloudwatch service.
2. We have disabled ssl upon USE_SSL
3. The spawned container need to communicate with docker in order to run the lambda. For this we need to define:
   • LAMBDA_EXECUTOR as docker
   • Map the docker socket file inside the container via a bind-mount. It is achieved via putting /var/run/docker.sock:/var/run/docker.sock upon volumes section.

Localstack is available thorugh http://localhost:4566 as defined upon:

 ports: - "4566:4566" 

Download Lambda code

Lambda code is usually stored as a .zip file, which contains both the application code and its dependencies:

Detect lambda entrypoint

Lambdas are node.js or python scripts, usually is is a function inside a specific file.

For example, in a Python Lambda, the entrypoint is often defined as file_name.handler_function:

In this case, AWS starts execution from the lambda_handler method inside lambda_function.py:

Identifying the entrypoint is crucial because it helps determine dependencies and setup.

Installing the Required Python or Node.js Version

AWS Lambdas run with specific Python (or Node.js) runtimes (e.g., Python 3.9, 3.10).

You can check which runtime is being used under the Code tab in the AWS Lambda console:

Organizing Code and Dependencies:

Once you’ve downloaded the Lambda .zip, unzip it into a working directory. I recommend creating two folders:

1. Original Lambda → The raw contents of the .zip. (for example my-lambda-orig
2. Clean Repo → A structured project where you separate code from dependencies. (for example my-lambda). Thefollowing stricture is reccomended:
   • src -> containing lambda source code
   • tools -> for usesfull tools

Why separate them?
AWS often packages Lambdas with both source code and dependencies. Instead, you’ll want to manage dependencies with pip (and later, requirements.txt/poetry/pipenv).

Here’s the process:

1. Copy the main Lambda file (e.g., lambda_function.py) into ./src.
2. Inspect the start of the Lambda file (e.g., lambda_function.py). There you can distinguish which files are needed ans which are 3rd party libraries. Any non 3rd party library copy them upon src folder of the clean repo.
3. Using the name you can detect which ones are 3rd-party and which ones are required. It is a good idea to keep a list with 3rd party libraries as well.
4. Copy only the application code (your .py files) into the clean repo.

A common Library is the boto one which is used for interfacing with AWS itself.

Once seperate the code init git and fo a first commit:

git init git commit -m "Cleaned Code" 

Create a virtualenv

We need to have a list with lambda dependencies for that we ned to initialize a virtual env:

python -m venv ./.venv echo ".env" > .gitignore git add .gitignore git commit -m "Ignoring virtualenv" source .venv/bin/activate 

Then using pip install we are installing the nesesary dependencies, for example for boto we run:

pip install boto3 

Once depencendies installed we need to keep them into a list for that we need to run:

pip freeze > ./requirements.txt git add requirements.txt git commit -m "Added Dependencies" 

Now every time we need to install them we can do:

pip -r requirements.txt 

Detect environmental variables

Lambas also retrieve settings externally from environmental variables. For pythonk lambdas scan for strings such as os.environ.get(, the parameter passed contains the environemtnal variable required.

Deploy upon localstack

Now that we have code sorted out we can create a lambda into localstack. For this we need to package the lambda as a new zip:

 ## Package lambda mkdir package echo "package" >> .gitignore git add .gitignore git commit -m "Ignoring support folder for packaging" # Copy code cp -r ./src package/ # Install dependencies upon package folder pip install -r requirements.txt -t package # Zip package cd package zip -r9 function.zip cd ../ # Package folder is temporary rm -rf ./package 

The commands above do 2 things:

• Create a intermediate folder where code is staged, the folder is named package. This folder and it should be ignored from git.
• Copy all lambda source upon package folder. keep in mind we need to copy the contents of src and not the src itself.
• Install all dependencies at package folder. This is achieved via providing the folder name upon -t parameter.

The method described above can also be used for packaging the lambda for production as well. The zip file can be uploaded, thus replacing the lambda code.

Deploy on locakstack

For this we can run:

 aws --endpoint-url=http://localhost:4566 lambda create-function\ --function-name "my-function" \ --runtime "python3.9" \ --role "arn:aws:iam::000000000000:role/lambda-role" \ --handler "python3.9" \ --zip-file "fileb://function.zip" \ --environment Variables="{VAR1=\"Hello\"}" 

This command above creates the function my-function this is the lambda function we need to create. At the command above we need to pay attention upon some arguments:

• --endpoint-url In our case we need to provide it with the localstack setup. If not it will try to create the lambda upon actual aws. In out case this si not desirable
• --fucntion-name a distinctive function name. Each lambda should have its own
• --role Keeping the value arn:aws:iam::000000000000:role/lambda-role is fine on every time tyou crreatr a new lambda. Localstack does not check for roles but we need to place this in order to avoid any unwanted behaviour.
• --handler: Place the one that actual lambda upon aws uses. Here we define what would be used for lambda execution.
• --zip-file The file we created containing code + dependencies.
• --environment If no environmental variables used, ommit it. The value is a string encloded in {} having this format (do not confuse it with json):

 {VARIABLE_NAME=VALUE,VARIABLE_NAME2=VALUE2} 

Develop and redeploy

During development if a changes is needed you need to re-package the lambda. Re-creating is not needed unless localstack container is terminated.

Updating the code is done like this:

aws --endpoint-url=http://localhost:4566 lambda update-function-code --function-name "my-function" --zip-file "fileb://function.zip" 

Whilst if environmental variables need to change you can do it like this:

aws --endpoint-url=$ENDPOINT lambda update-function-configuration \ --function-name "my-function" \ --environment Variables="{VARIABLE_NAME=VALUE,VARIABLE_NAME2=VALUE2}" 

Misc tips

If an extra dependency is needed is it a good idea to enable the generated virtual environment, then install it via pip. Then update the code upon localstack as mentioned above.