Skip to content

A tool for building lambda function images or zips via Docker

License

Notifications You must be signed in to change notification settings

dokku/lambda-builder

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

lambda-builder

A tool for building lamda functions into uploadable zip files via Docker based on work from @lambci and @mLupine.

Why?

I don't want to go through the motions of figuring out the correct way to build my app for AWS. I suspect there are others out there who feel the same.

Dependencies

  • The docker binary
  • Golang 1.7+

Building

# substitute the version number as desired
go build -ldflags "-X main.Version=0.5.0"

Usage

Usage: lambda-builder [--version] [--help] <command> [<args>]

Available commands are:
    build      Builds a lambda function
    version    Return the version of the binary

Building an app

To build an app:

cd path/to/app

# will write a lambda.zip in the current working directory
lambda-builder build

Alternatively, a given path can be specified via the --working-directory flag:

# will write a lambda.zip in the specified path
lambda-builder build --working-directory path/to/app

Custom environment variables can be supplied for the build environment by specifying one or more --build-env flags. The --build-env flag takes KEY=VALUE pairs.

# the build step will have access to both the --build-env pairs
lambda-builder build --build-env KEY=VALUE --build-env ANOTHER_KEY=some-value

A builder can be chosen by a flag. Note that while a builder may be selected, the detection for that builder must still pass in order for the build to succeed.

lambda-builder build --generate-image --builder dotnet

Building an image

A docker image can be produced from the generated artifact by specifying the --generate-image flag. This also allows for multiple --label flags as well as specifying a single image tag via either -t or --tag:

# will write a lambda.zip in the specified path
# and generate a docker image named `lambda-builder/$APP:latest`
# where $APP is the last portion of the working directory
lambda-builder build --generate-image

# adds the labels com.example/key=value and com.example/another-key=value
lambda-builder build --generate-image --label com.example/key=value --label com.example/another-key=value

# tags the image as app/awesome:1234
lambda-builder build --generate-image --tag app/awesome:1234

By default, any web process started by the built image starts on port 9001. This can be overriden via the --port environment variable.

# build the image and ensure it starts on port 5000 by default
lambda-builder build --generate-image --port 5000

Custom environment variables can be supplied for the built image by specifying one or more --image-env flags. The --image-env flag takes KEY=VALUE pairs.

# the built image will have `ENV` directives corresponding to the values specified by `--image-env`
lambda-builder build --generate-image --image-env KEY=VALUE --image-env ANOTHER_KEY=some-value

The build-image and run-image can also be specified as flags:

lambda-builder build --generate-image --build-image "mlupin/docker-lambda:dotnetcore3.1-build" --run-image "mlupin/docker-lambda:dotnetcore3.1"

A generated image can be run locally with the following line:

# run the container and ensure it stays open
# replace `$APP` with your folder name
docker run --rm -it -e DOCKER_LAMBDA_STAY_OPEN=1 -p 9001:9001 "lambda-builder/$APP:latest"

# invoke it using the awscli (v2)
# note that the function name in this example is `function.handler`
aws lambda invoke --endpoint http://localhost:9001 --no-sign-request --function-name function.handler --payload '{}' --cli-binary-format raw-in-base64-out output.json

# invoke it via curl
curl -d '{}' http://localhost:9001/2015-03-31/functions/function.handler/invocations

# the function can also be invoked directly from a container if desired
docker run --rm "lambda-builder/$APP:latest" function.handler '{"name": "World"}'

Generating a Procfile

A Procfile can be written to the working directory by specifying the --write-procfile flag. This file will not be written if one already exists in the working directory. If an image is being built, the detected handler will also be injected into the build context and used as the default CMD for the image. The contents of the Procfile are a web process type and a detected handler.

# writes out a procfile
lambda-builder build --write-procfile

A --handler flag can be specified with a custom handler to override the one detected.

# override with a custom handler
lambda-builder build --write-procfile --handler foo_file.bar_func

How does it work

Internally, lambda-builder detects a given language and builds the app according to the script specified by the detected builder within a disposablecontainer environment emulating AWS Lambda. If a builder is not detected, the build will fail. The following languages are supported:

  • dotnet
    • default build image: mlupin/docker-lambda:dotnet6-build
    • requirement: Function.cs
    • runtimes:
      • dotnet6
      • dotnetcore3.1
  • go
    • default build image:
      • With go.mod: golang:1.22-bookworm
      • Without go.mod: golang:1.17-buster
    • requirement: go.mod or main.go
    • runtimes:
      • provided.al2
  • nodejs
    • default build image: mlupin/docker-lambda:nodejs14.x-build
    • requirement: package-lock.json
    • runtimes:
      • nodejs12.x
      • nodejs14.x
  • python
    • default build image: mlupin/docker-lambda:python3.9-build
    • requirement: requirements.txt, poetry.lock, or Pipfile.lock
    • notes: Autodetects the python version from poetry.lock, Pipfile.lock, or runtime.txt
    • runtimes:
      • python3.8
      • python3.9
  • ruby
    • default build image: mlupin/docker-lambda:ruby2.7-build
    • requirement: Gemfile.lock
    • runtimes:
      • ruby2.7

All builders support both pre (run before the app is compiled) and post (run after the app is compiled but before it is compressed into a lambda.zip file) compile hooks in the form of bin/pre_compile and bin/post_compile. These can be shell scripts or executables.

When the app is built, a lambda.zip will be produced in the specified working directory. The resulting lambda.zip can be uploaded to S3 and used within a Lambda function.

Both the builder, build image environment, and the run image environment can be overriden in an optional lambda.yml file in the specified working directory.

lambda.yml

The following a short description of the lambda.yml format.

---
build_image: mlupin/docker-lambda:dotnetcore3.1-build
builder: dotnet
run_image: mlupin/docker-lambda:dotnetcore3.1
  • build_image: A docker image that is accessible by the docker daemon. The build_image should be based on an existing Lambda image - builders may fail if they cannot run within the specified build_image. The build will fail if the image is inaccessible by the docker daemon.
  • builder: The name of a builder. This may be used if multiple builders match and a specific builder is desired. If an invalid builder is specified, the build will fail.
  • run_image: A docker image that is accessible by the docker daemon. The run_image should be based on an existing Lambda image - built images may fail to start if they are not compatible with the produced artifact. The generation of the run iage will fail if the image is inaccessible by the docker daemon.

Deploying

The lambda.zip file can be directly uploaded to a lambda function and used as is by specifying the correct runtime. See the test.bats files in any of the test examples for more info on how to perform this with the awscli (v2).

Examples

See the tests directory for examples on how to use this project.