Terraform module, which creates almost all supported AWS Lambda resources as well as taking care of building and packaging of required Lambda dependencies for functions and layers.

This Terraform module is the part of serverless.tf framework, which aims to simplify all operations when working with the serverless in Terraform:

1. Build and install dependencies - read more. Requires Python 3.6 or newer.
2. Create, store, and use deployment packages - read more.
3. Create, update, and publish AWS Lambda Function and Lambda Layer - see usage.
4. Create static and dynamic aliases for AWS Lambda Function - see usage, see modules/alias.
5. Do complex deployments (eg, rolling, canary, rollbacks, triggers) - read more, see modules/deploy.
6. Use AWS SAM CLI to test Lambda Function - read more.

• Build dependencies for your Lambda Function and Layer.
• Support builds locally and in Docker (with or without SSH agent support for private builds).
• Create deployment package or deploy existing (previously built package) from local, from S3, from URL, or from AWS ECR repository.
• Store deployment packages locally or in the S3 bucket.
• Support almost all features of Lambda resources (function, layer, alias, etc.)
• Lambda@Edge
• Conditional creation for many types of resources.
• Control execution of nearly any step in the process - build, package, store package, deploy, update.
• Control nearly all aspects of Lambda resources (provisioned concurrency, VPC, EFS, dead-letter notification, tracing, async events, event source mapping, IAM role, IAM policies, and more).
• Support integration with other serverless.tf modules like HTTP API Gateway (see examples there).

module "lambda_function" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda1"
  description   = "My awesome lambda function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  source_path = "../src/lambda-function1"

  tags = {
    Name = "my-lambda1"
  }
}

module "lambda_function" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "lambda-with-layer"
  description   = "My awesome lambda function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"
  publish       = true

  source_path = "../src/lambda-function1"

  store_on_s3 = true
  s3_bucket   = "my-bucket-id-with-lambda-builds"

  layers = [
    module.lambda_layer_s3.lambda_layer_arn,
  ]

  environment_variables = {
    Serverless = "Terraform"
  }

  tags = {
    Module = "lambda-with-layer"
  }
}

module "lambda_layer_s3" {
  source = "terraform-aws-modules/lambda/aws"

  create_layer = true

  layer_name          = "lambda-layer-s3"
  description         = "My amazing lambda layer (deployed from S3)"
  compatible_runtimes = ["python3.8"]

  source_path = "../src/lambda-layer"

  store_on_s3 = true
  s3_bucket   = "my-bucket-id-with-lambda-builds"
}

module "lambda_function_existing_package_local" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda-existing-package-local"
  description   = "My awesome lambda function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  create_package         = false
  local_existing_package = "../existing_package.zip"
}

If you want to manage function code and infrastructure resources (such as IAM permissions, policies, events, etc) in separate flows (e.g., different repositories, teams, CI/CD pipelines).

Disable source code tracking to turn off deployments (and rollbacks) using the module by setting ignore_source_code_hash = true and deploy a dummy function.

When the infrastructure and the dummy function is deployed, you can use external tool to update the source code of the function (eg, using AWS CLI) and keep using this module via Terraform to manage the infrastructure.

Be aware that changes in local_existing_package value may trigger deployment via Terraform.

module "lambda_function_externally_managed_package" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda-externally-managed-package"
  description   = "My lambda function code is deployed separately"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  create_package         = false
  local_existing_package = "./lambda_functions/code.zip"

  ignore_source_code_hash = true
}

Note that this module does not copy prebuilt packages into S3 bucket. This module can only store packages it builds locally and in S3 bucket.

locals {
  my_function_source = "../path/to/package.zip"
}

resource "aws_s3_bucket" "builds" {
  bucket = "my-builds"
  acl    = "private"
}

resource "aws_s3_object" "my_function" {
  bucket = aws_s3_bucket.builds.id
  key    = "${filemd5(local.my_function_source)}.zip"
  source = local.my_function_source
}

module "lambda_function_existing_package_s3" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda-existing-package-local"
  description   = "My awesome lambda function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  create_package      = false
  s3_existing_package = {
    bucket = aws_s3_bucket.builds.id
    key    = aws_s3_object.my_function.id
  }
}

module "lambda_function_container_image" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda-existing-package-local"
  description   = "My awesome lambda function"

  create_package = false

  image_uri    = "132367819851.dkr.ecr.eu-west-1.amazonaws.com/complete-cow:1.0"
  package_type = "Image"
}

module "lambda_layer_local" {
  source = "terraform-aws-modules/lambda/aws"

  create_layer = true

  layer_name          = "my-layer-local"
  description         = "My amazing lambda layer (deployed from local)"
  compatible_runtimes = ["python3.8"]

  source_path = "../fixtures/python3.8-app1"
}

module "lambda_layer_s3" {
  source = "terraform-aws-modules/lambda/aws"

  create_layer = true

  layer_name          = "my-layer-s3"
  description         = "My amazing lambda layer (deployed from S3)"
  compatible_runtimes = ["python3.8"]

  source_path = "../fixtures/python3.8-app1"

  store_on_s3 = true
  s3_bucket   = "my-bucket-id-with-lambda-builds"
}

Make sure, you deploy Lambda@Edge functions into US East (N. Virginia) region (us-east-1). See Requirements and Restrictions on Lambda Functions.

module "lambda_at_edge" {
  source = "terraform-aws-modules/lambda/aws"

  lambda_at_edge = true

  function_name = "my-lambda-at-edge"
  description   = "My awesome lambda@edge function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  source_path = "../fixtures/python3.8-app1"

  tags = {
    Module = "lambda-at-edge"
  }
}

module "lambda_function_in_vpc" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda-in-vpc"
  description   = "My awesome lambda function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  source_path = "../fixtures/python3.8-app1"

  vpc_subnet_ids         = module.vpc.intra_subnets
  vpc_security_group_ids = [module.vpc.default_security_group_id]
  attach_network_policy = true
}

module "vpc" {
  source = "terraform-aws-modules/vpc/aws"

  name = "my-vpc"
  cidr = "10.10.0.0/16"

  # Specify at least one of: intra_subnets, private_subnets, or public_subnets
  azs           = ["eu-west-1a", "eu-west-1b", "eu-west-1c"]
  intra_subnets = ["10.10.101.0/24", "10.10.102.0/24", "10.10.103.0/24"]
}

There are 6 supported ways to attach IAM policies to IAM role used by Lambda Function:

1. policy_json - JSON string or heredoc, when attach_policy_json = true.
2. policy_jsons - List of JSON strings or heredoc, when attach_policy_jsons = true and number_of_policy_jsons > 0.
3. policy - ARN of existing IAM policy, when attach_policy = true.
4. policies - List of ARNs of existing IAM policies, when attach_policies = true and number_of_policies > 0.
5. policy_statements - Map of maps to define IAM statements which will be generated as IAM policy. Requires attach_policy_statements = true. See examples/complete for more information.
6. assume_role_policy_statements - Map of maps to define IAM statements which will be generated as IAM policy for assuming Lambda Function role (trust relationship). See examples/complete for more information.

Lambda Permissions should be specified to allow certain resources to invoke Lambda Function.

module "lambda_function" {
  source = "terraform-aws-modules/lambda/aws"

  # ...omitted for brevity

  allowed_triggers = {
    Config = {
      principal        = "config.amazonaws.com"
      principal_org_id = "o-abcdefghij"
    }
    APIGatewayAny = {
      service    = "apigateway"
      source_arn = "arn:aws:execute-api:eu-west-1:135367859851:aqnku8akd0/*/*/*"
    },
    APIGatewayDevPost = {
      service    = "apigateway"
      source_arn = "arn:aws:execute-api:eu-west-1:135367859851:aqnku8akd0/dev/POST/*"
    },
    OneRule = {
      principal  = "events.amazonaws.com"
      source_arn = "arn:aws:events:eu-west-1:135367859851:rule/RunDaily"
    }
  }
}

Sometimes you need to have a way to create resources conditionally but Terraform does not allow usage of count inside module block, so the solution is to specify create arguments.

module "lambda" {
  source = "terraform-aws-modules/lambda/aws"

  create = false # to disable all resources

  create_package  = false  # to control build package process
  create_function = false  # to control creation of the Lambda Function and related resources
  create_layer    = false  # to control creation of the Lambda Layer and related resources
  create_role     = false  # to control creation of the IAM role and policies required for Lambda Function

  attach_cloudwatch_logs_policy = false
  attach_dead_letter_policy     = false
  attach_network_policy         = false
  attach_tracing_policy         = false
  attach_async_event_policy     = false

  # ... omitted
}

This is one of the most complicated part done by the module and normally you don't have to know internals.

package.py is Python script which does it. Make sure, Python 3.6 or newer is installed. The main functions of the script are to generate a filename of zip-archive based on the content of the files, verify if zip-archive has been already created, and create zip-archive only when it is necessary (during apply, not plan).

Hash of zip-archive created with the same content of the files is always identical which prevents unnecessary force-updates of the Lambda resources unless content modifies. If you need to have different filenames for the same content you can specify extra string argument hash_extra.

When calling this module multiple times in one execution to create packages with the same source_path, zip-archives will be corrupted due to concurrent writes into the same file. There are two solutions - set different values for hash_extra to create different archives, or create package once outside (using this module) and then pass local_existing_package argument to create other Lambda resources.

Building and packaging has been historically hard to debug (especially with Terraform), so we made an effort to make it easier for user to see debug info. There are 3 different debug levels: DEBUG - to see only what is happening during planning phase and how a zip file content filtering in case of applied patterns, DEBUG2 - to see more logging output, DEBUG3 - to see all logging values, DUMP_ENV - to see all logging values and env variables (be careful sharing your env variables as they may contain secrets!).

User can specify debug level like this:

export TF_LAMBDA_PACKAGE_LOG_LEVEL=DEBUG2
terraform apply

User can enable comments in heredoc strings in patterns which can be helpful in some situations. To do this set this environment variable:

export TF_LAMBDA_PACKAGE_PATTERN_COMMENTS=true
terraform apply

You can specify source_path in a variety of ways to achieve desired flexibility when building deployment packages locally or in Docker. You can use absolute or relative paths. If you have placed terraform files in subdirectories, note that relative paths are specified from the directory where terraform plan is run and not the location of your terraform file.

Note that, when building locally, files are not copying anywhere from the source directories when making packages, we use fast Python regular expressions to find matching files and directories, which makes packaging very fast and easy to understand.

When source_path is set to a string, the content of that path will be used to create deployment package as-is:

source_path = "src/function1"

When source_path is set to a list of directories the content of each will be taken and one archive will be created.

This is the most complete way of creating a deployment package from multiple sources with multiple dependencies. This example is showing some of the available options (see examples/build-package for more):

source_path = [
  "src/main-source",
  "src/another-source/index.py",
  {
    path     = "src/function1-dep",
    patterns = [
      "!.*/.*\\.txt", # Skip all txt files recursively
    ]
  }, {
    path             = "src/python3.8-app1",
    pip_requirements = true,
    pip_tmp_dir      = "/tmp/dir/location"
    prefix_in_zip    = "foo/bar1",
  }, {
    path             = "src/python3.8-app2",
    pip_requirements = "requirements-large.txt",
    patterns = [
      "!vendor/colorful-0.5.4.dist-info/RECORD",
      "!vendor/colorful-.+.dist-info/.*",
      "!vendor/colorful/__pycache__/?.*",
    ]
  }, {
    path             = "src/nodejs14.x-app1",
    npm_requirements = true,
    npm_tmp_dir      = "/tmp/dir/location"
    prefix_in_zip    = "foo/bar1",
  }, {
    path     = "src/python3.8-app3",
    commands = [
      "npm install",
      ":zip"
    ],
    patterns = [
      "!.*/.*\\.txt",    # Skip all txt files recursively
      "node_modules/.+", # Include all node_modules
    ],
  }, {
    path     = "src/python3.8-app3",
    commands = ["go build"],
    patterns = <<END
      bin/.*
      abc/def/.*
    END
  }
]

Few notes:

• If you specify a source path as a string that references a folder and the runtime begins with python or nodejs, the build process will automatically build python and nodejs dependencies if requirements.txt or package.json file will be found in the source folder. If you want to customize this behavior, please use the object notation as explained below.
• All arguments except path are optional.
• patterns - List of Python regex filenames should satisfy. Default value is "include everything" which is equal to patterns = [".*"]. This can also be specified as multiline heredoc string (no comments allowed). Some examples of valid patterns:
• If you use the commands option and chain multiple commands, only the exit code of last command will be checked for success. If you prefer to fail fast, start the commands with the bash option set -e or powershell option $ErrorActionPreference="Stop"

    !.*/.*\.txt        # Filter all txt files recursively
    node_modules/.*    # Include empty dir or with a content if it exists
    node_modules/.+    # Include full non empty node_modules dir with its content
    node_modules/      # Include node_modules itself without its content
                       # It's also a way to include an empty dir if it exists
    node_modules       # Include a file or an existing dir only

    !abc/.*            # Filter out everything in an abc folder
    abc/def/.*         # Re-include everything in abc/def sub folder
    !abc/def/hgk/.*    # Filter out again in abc/def/hgk sub folder

• commands - List of commands to run. If specified, this argument overrides pip_requirements and npm_requirements.
  • :zip [source] [destination] is a special command which creates content of current working directory (first argument) and places it inside of path (second argument).
• pip_requirements - Controls whether to execute pip install. Set to false to disable this feature, true to run pip install with requirements.txt found in path. Or set to another filename which you want to use instead. When source_path is passed as a string containing a path (and not a list of maps), and requirements.txt is present, pip install is automatically executed.
• pip_tmp_dir - Set the base directory to make the temporary directory for pip installs. Can be useful for Docker in Docker builds.
• poetry_install - Controls whether to execute poetry export and pip install. Set to false to disable this feature, true to run poetry export with pyproject.toml and poetry.lock found in path. When source_path is passed as a string containing a path (and not a list of maps), and pyproject.toml with a build system poetry is present, poetry export and pip install are automatically executed.
• npm_requirements - Controls whether to execute npm install. Set to false to disable this feature, true to run npm install with package.json found in path. Or set to another filename which you want to use instead.
• npm_tmp_dir - Set the base directory to make the temporary directory for npm installs. Can be useful for Docker in Docker builds.
• prefix_in_zip - If specified, will be used as a prefix inside zip-archive. By default, everything installs into the root of zip-archive.

If your Lambda Function or Layer uses some dependencies you can build them in Docker and have them included into deployment package. Here is how you can do it:

build_in_docker   = true
docker_file       = "src/python3.8-app1/docker/Dockerfile"
docker_build_root = "src/python3.8-app1/docker"
docker_image      = "public.ecr.aws/sam/build-python3.8"
runtime           = "python3.8"    # Setting runtime is required when building package in Docker and Lambda Layer resource.

Using this module you can install dependencies from private hosts. To do this, you need for forward SSH agent:

docker_with_ssh_agent = true

Note that by default, the docker_image used comes from the registry public.ecr.aws/sam/, and will be based on the runtime that you specify. In other words, if you specify a runtime of python3.8 and do not specify docker_image, then the docker_image will resolve to public.ecr.aws/sam/build-python3.8. This ensures that by default the runtime is available in the docker container.

If you override docker_image, be sure to keep the image in sync with your runtime. During the plan phase, when using docker, there is no check that the runtime is available to build the package. That means that if you use an image that does not have the runtime, the plan will still succeed, but then the apply will fail.

To add flexibility when building in docker, you can pass any number of additional options that you require (see Docker run reference for available options):

  docker_additional_options = [
        "-e", "MY_ENV_VAR='My environment variable value'",
        "-v", "/local:/docker-vol",
  ]

To override the docker entrypoint when building in docker, set docker_entrypoint:

  docker_entrypoint = "/entrypoint/entrypoint.sh"

The entrypoint must map to a path within your container, so you need to either build your own image that contains the entrypoint or map it to a file on the host by mounting a volume (see Passing additional Docker options).

By default, this module creates deployment package and uses it to create or update Lambda Function or Lambda Layer.

Sometimes, you may want to separate build of deployment package (eg, to compile and install dependencies) from the deployment of a package into two separate steps.

When creating archive locally outside of this module you need to set create_package = false and then argument local_existing_package = "existing_package.zip". Alternatively, you may prefer to keep your deployment packages into S3 bucket and provide a reference to them like this:

  create_package      = false
  s3_existing_package = {
    bucket = "my-bucket-with-lambda-builds"
    key    = "existing_package.zip"
  }

This can be implemented in two steps: download file locally using CURL, and pass path to deployment package as local_existing_package argument.

locals {
  package_url = "https://raw.githubusercontent.com/terraform-aws-modules/terraform-aws-lambda/master/examples/fixtures/python3.8-zip/existing_package.zip"
  downloaded  = "downloaded_package_${md5(local.package_url)}.zip"
}

resource "null_resource" "download_package" {
  triggers = {
    downloaded = local.downloaded
  }

  provisioner "local-exec" {
    command = "curl -L -o ${local.downloaded} ${local.package_url}"
  }
}

data "null_data_source" "downloaded_package" {
  inputs = {
    id       = null_resource.download_package.id
    filename = local.downloaded
  }
}

module "lambda_function_existing_package_from_remote_url" {
  source = "terraform-aws-modules/lambda/aws"

  function_name = "my-lambda-existing-package-local"
  description   = "My awesome lambda function"
  handler       = "index.lambda_handler"
  runtime       = "python3.8"

  create_package         = false
  local_existing_package = data.null_data_source.downloaded_package.outputs["filename"]
}

AWS SAM CLI is an open source tool that help the developers to initiate, build, test, and deploy serverless applications. SAM CLI tool supports Terraform applications.

SAM CLI provides two ways of testing: local testing and testing on-cloud (Accelerate).

Using SAM CLI, you can invoke the lambda functions defined in the terraform application locally using the sam local invoke command, providing the function terraform address, or function name, and to set the hook-name to terraform to tell SAM CLI that the underlying project is a terraform application.

You can execute the sam local invoke command from your terraform application root directory as following:

sam local invoke --hook-name terraform module.hello_world_function.aws_lambda_function.this[0]

You can also pass an event to your lambda function, or overwrite its environment variables. Check here for more information.

You can also invoke your lambda function in debugging mode, and step-through your lambda function source code locally in your preferred editor. Check here for more information.

You can use AWS SAM CLI to quickly test your application on your AWS development account. Using SAM Accelerate, you will be able to develop your lambda functions locally, and once you save your updates, SAM CLI will update your development account with the updated Lambda functions. So, you can test it on cloud, and if there is any bug, you can quickly update the code, and SAM CLI will take care of pushing it to the cloud. Check here for more information about SAM Accelerate.

You can execute the sam sync command from your terraform application root directory as following:

sam sync --hook-name terraform --watch

Typically, Lambda Function resource updates when source code changes. If publish = true is specified a new Lambda Function version will also be created.

Published Lambda Function can be invoked using either by version number or using $LATEST. This is the simplest way of deployment which does not required any additional tool or service.

In order to do controlled deployments (rolling, canary, rollbacks) of Lambda Functions we need to use Lambda Function aliases.

In simple terms, Lambda alias is like a pointer to either one version of Lambda Function (when deployment complete), or to two weighted versions of Lambda Function (during rolling or canary deployment).

One Lambda Function can be used in multiple aliases. Using aliases gives large control of which version deployed when having multiple environments.

There is alias module, which simplifies working with alias (create, manage configurations, updates, etc). See examples/alias for various use-cases how aliases can be configured and used.

There is deploy module, which creates required resources to do deployments using AWS CodeDeploy. It also creates the deployment, and wait for completion. See examples/deploy for complete end-to-end build/update/deploy process.

Terraform Cloud, Terraform Enterprise, and many other SaaS for running Terraform do not have Python pre-installed on the workers. You will need to provide an alternative Docker image with Python installed to be able to use this module there.

Q1: Why deployment package not recreating every time I change something? Or why deployment package is being recreated every time but content has not been changed?

Answer: There can be several reasons related to concurrent executions, or to content hash. Sometimes, changes has happened inside of dependency which is not used in calculating content hash. Or multiple packages are creating at the same time from the same sources. You can force it by setting value of hash_extra to distinct values.

Q2: How to force recreate deployment package?

Answer: Delete an existing zip-archive from builds directory, or make a change in your source code. If there is no zip-archive for the current content hash, it will be recreated during terraform apply.

Q3: null_resource.archive[0] must be replaced

Answer: This probably mean that zip-archive has been deployed, but is currently absent locally, and it has to be recreated locally. When you run into this issue during CI/CD process (where workspace is clean) or from multiple workspaces, you can set environment variable TF_RECREATE_MISSING_LAMBDA_PACKAGE=false or pass recreate_missing_package = false as a parameter to the module and run terraform apply. Alternatively, you can pass trigger_on_package_timestamp = false as a parameter to ignore the file timestamp when deciding to create the archive or not.

Q4: What does this error mean - "We currently do not support adding policies for $LATEST." ?

Answer: When the Lambda function is created with publish = true the new version is automatically increased and a qualified identifier (version number) becomes available and will be used when setting Lambda permissions.

When publish = false (default), only unqualified identifier ($LATEST) is available which leads to the error.

The solution is to either disable the creation of Lambda permissions for the current version by setting create_current_version_allowed_triggers = false, or to enable publish of Lambda function (publish = true).

1. Creation of Lambda Functions and Lambda Layers is very similar and both support the same features (building from source path, using existing package, storing package locally or on S3)
2. Check out this Awesome list of AWS Lambda Layers

• Complete - Create Lambda resources in various combinations with all supported features.
• Container Image - Create a Docker image with a platform specified in the Dockerfile (using docker provider), push it to AWS ECR, and create Lambda function from it.
• Build and Package - Build and create deployment packages in various ways.
• Alias - Create static and dynamic aliases in various ways.
• Deploy - Complete end-to-end build/update/deploy process using AWS CodeDeploy.
• Async Invocations - Create Lambda Function with async event configuration (with SQS, SNS, and EventBridge integration).
• With VPC - Create Lambda Function with VPC.
• With VPC and VPC Endpoint for S3 - Create Lambda Function with VPC and VPC Endpoint for S3.
• With EFS - Create Lambda Function with Elastic File System attached (Terraform 0.13+ is recommended).
• Multiple regions - Create the same Lambda Function in multiple regions with non-conflicting IAM roles and policies.
• Event Source Mapping - Create Lambda Function with event source mapping configuration (SQS, DynamoDB, Amazon MQ, and Kinesis).
• Triggers - Create Lambda Function with some triggers (eg, Cloudwatch Events, EventBridge).
• Code Signing - Create Lambda Function with code signing configuration.
• Simple CI/CD - Create Lambda Function as if it runs on CI/CD platform where builds directory is often absent.

• 1Mill/serverless-tf-examples

No modules.

During development involving modifying python files, use tox to run unit tests:

tox

This will try to run unit tests which each supported python version, reporting errors for python versions which are not installed locally.

If you only want to test against your main python version:

tox -e py

You can also pass additional positional arguments to pytest which is used to run test, e.g. to make it verbose:

tox -e py -- -vvv

Module managed by Anton Babenko. Check out serverless.tf to learn more about doing serverless with Terraform.

Please reach out to Betajob if you are looking for commercial support for your Terraform, AWS, or serverless project.

Apache 2 Licensed. See LICENSE for full details.

• Russia has illegally annexed Crimea in 2014 and brought the war in Donbas followed by full-scale invasion of Ukraine in 2022.
• Russia has brought sorrow and devastations to millions of Ukrainians, killed hundreds of innocent people, damaged thousands of buildings, and forced several million people to flee.
• Putin khuylo!