AWS Compute Blog

Better together: AWS SAM CLI and HashiCorp Terraform

This post is written by Suresh Poopandi, Senior Solutions Architect and Seb Kasprzak, Senior Solutions Architect.

Today, AWS is announcing the public preview of AWS Serverless Application Model CLI (AWS SAM CLI) support for local development, testing, and debugging of serverless applications defined using HashiCorp Terraform configuration.

AWS SAM and Terraform are open-source frameworks for building applications using infrastructure as code (IaC). Both frameworks allow building, changing, and managing cloud infrastructure in a repeatable way by defining resource configurations.

Previously, you could use the AWS SAM CLI to build, test, and debug applications defined by AWS SAM templates or through the AWS Cloud Development Kit (CDK). With this preview release, you can also use AWS SAM CLI to test and debug serverless applications defined using Terraform configurations.

Walkthrough of Terraform support

This blog post contains a sample Terraform template, which shows how developers can use AWS SAM CLI to build locally, test, and debug AWS Lambda functions defined in Terraform. This sample application has a Lambda function that stores a book review score and review text in an Amazon DynamoDB table. An Amazon API Gateway book review API uses Lambda proxy integration to invoke the book review Lambda function.

Demo application architecture

Demo application architecture


Before running this example:

  • Install the AWS CLI.
    • Configure with valid AWS credentials.
    • Note that AWS CLI now requires Python runtime.
  • Install HashiCorp Terraform.
  • Install the AWS SAM CLI.
  • Install Docker (required to run AWS Lambda function locally).

Since Terraform support is currently in public preview, you must provide a –beta-features flag while executing AWS SAM commands. Alternatively, set this flag in samconfig.toml file by adding beta_features=”true”.

Deploying the example application

This Lambda function interacts with DynamoDB. For the example to work, it requires an existing DynamoDB table in an AWS account. Deploying this creates all the required resources for local testing and debugging of the Lambda function.

To deploy:

  1. Clone the aws-sam-terraform-examples repository locally:
    git clone
  2. Change to the project directory:
    cd aws-sam-terraform-examples/zip_based_lambda_functions/api-lambda-dynamodb-example/

    Terraform must store the state of the infrastructure and configuration it creates. Terraform uses this state to map cloud resources to configuration and track changes. This example uses a local backend to store the state file on the local filesystem.

  3. Open the file and review its contents. Locate the backend section of the code, updating the region field with the target deployment Region of this sample solution:
    provider “aws” {
        region = “<AWS region>” # e.g. us-east-1
  4. Initialize a working directory containing Terraform configuration files:
    terraform init
  5. Deploy the application using Terraform CLI. When prompted by “Do you want to perform these actions?”, enter Yes.
    terraform apply

Terraform deploys the application, as shown in the terminal output.

Terminal output

Terminal output

After completing the deployment process, the AWS account is ready for use by the Lambda function with all the required resources.

Terraform Configuration for local testing

Lambda functions require application dependencies bundled together with function code as a deployment package (typically a .zip file) to run correctly. Terraform natively does not create the deployment package and a separate build process handles this package creation.

This sample application uses Terraform’s null_resource and local-exec provisioner to trigger a build process script. This installs Python dependencies in a temporary folder and creates a .zip file with dependencies and function code. It contains this logic within the file of the example application.

To explain each code segment in more detail:

Terraform example

Terraform example

  1. aws_lambda_function: This sample defines a Lambda function resource. It contains properties such as environment variables (in this example, the DynamoDB table_id) and the depends_on argument, which creates the .zip package before deploying the Lambda function.

    Terraform example

    Terraform example

  2. null_resource: When the AWS SAM CLI build command runs, AWS SAM reviews Terraform code for any null_resource starting with sam_metadata_ and uses the information contained within this resource block to gather the location of the Lambda function source code and .zip package. This information allows the AWS SAM CLI to start the local execution of the Lambda function. This special resource should contain the following attributes:
    • resource_name: The Lambda function address as defined in the current module (aws_lambda_function.publish_book_review)
    • resource_type: Packaging type of the Lambda function (ZIP_LAMBDA_FUNCTION)
    • original_source_code: Location of Lambda function code
    • built_output_path: Location of .zip deployment package

Local testing

With the backend services now deployed, run local tests to see if everything is working. The locally running sample Lambda function interacts with the services deployed in the AWS account. Run the sam build to reflect the local sam testing environment with changes after each code update.

  1. Local Build: To create a local build of the Lambda function for testing, use the sam build command:
    sam build --hook-name terraform --beta-features
  2. Local invoke: The first test is to invoke the Lambda function with a mocked event payload from the API Gateway. These events are in the events directory. Run this command, passing in a mocked event:
    AWS_DEFAULT_REGION=<Your Region Name>
    sam local invoke aws_lambda_function.publish_book_review -e events/new-review.json --beta-features

    AWS SAM mounts the Lambda function runtime and code and runs it locally. The function makes a request to the DynamoDB table in the cloud to store the information provided via the API. It returns a 200 response code, signaling the successful completion of the function.

Local invoke from AWS CLI
Another test is to run a local emulation of the Lambda service using “sam local start-lambda” and invoke the function directly using AWS SDK or the AWS CLI. Start the local emulator with the following command:

sam local start-lambda
Terminal output

Terminal output

AWS SAM starts the emulator and exposes a local endpoint for the AWS CLI or a software development kit (SDK) to call. With the start-lambda command still running, run the following command to invoke this function locally with the AWS CLI:

aws lambda invoke --function-name aws_lambda_function.publish_book_review --endpoint-url response.json --cli-binary-format raw-in-base64-out --payload file://events/new-review.json

The AWS CLI invokes the local function and returns a status report of the service to the screen. The response from the function itself is in the response.json file. The window shows the following messages:

Invocation results

Invocation results

Debugging the Lambda function

Developers can use AWS SAM with a variety of AWS toolkits and debuggers to test and debug serverless applications locally. For example, developers can perform local step-through debugging of Lambda functions by setting breakpoints, inspecting variables, and running function code one line at a time.

The AWS Toolkit Integrated Development Environment (IDE) plugin provides the ability to perform many common debugging tasks, like setting breakpoints, inspecting variables, and running function code one line at a time. AWS Toolkits make it easier to develop, debug, and deploy serverless applications defined using AWS SAM. They provide an experience for building, testing, debugging, deploying, and invoking Lambda functions integrated into IDE. Refer to this link that lists common IDE/runtime combinations that support step-through debugging of AWS SAM applications.

Visual Studio Code keeps debugging configuration information in a launch.json file in a workspace .vscode folder. Here is a sample launch configuration file to debug Lambda code locally using AWS SAM and Visual Studio Code.

    "version": "0.2.0",
    "configurations": [
            "name": "Attach to SAM CLI",
            "type": "python",
            "request": "attach",
            "address": "localhost",
            "port": 9999,
            "localRoot": "${workspaceRoot}/sam-terraform/book-reviews",
            "remoteRoot": "/var/task",
            "protocol": "inspector",
            "stopOnEntry": false

After adding the launch configuration, start a debug session in the Visual Studio Code.

Step 1: Uncomment the following two lines in zip_based_lambda_functions/api-lambda-dynamodb-example/src/

Enable debugging in the Lambda function

Enable debugging in the Lambda function

Step 2: Run the Lambda function in the debug mode and wait for the Visual Studio Code to attach to this debugging session:

sam local invoke aws_lambda_function.publish_book_review -e events/new-review.json -d 9999

Step 3: Select the Run and Debug icon in the Activity Bar on the side of VS Code. In the Run and Debug view, select “Attach to SAM CLI” and choose Run.

For this example, set a breakpoint at the first line of lambda_handler. This breakpoint allows viewing the input data coming into the Lambda function. Also, it helps in debugging code issues before deploying to the AWS Cloud.

Debugging in then IDE

Debugging in then IDE

Lambda Terraform module

A community-supported Terraform module for lambda (terraform-aws-lambda) has added support for SAM metadata null_resource. When using the latest version of this module, AWS SAM CLI will automatically support local invocation of the Lambda function, without additional resource blocks required.


This blog post shows how to use the AWS SAM CLI together with HashiCorp Terraform to develop and test serverless applications in a local environment. With AWS SAM CLI’s support for HashiCorp Terraform, developers can now use the AWS SAM CLI to test their serverless functions locally while choosing their preferred infrastructure as code tooling.

For more information about the features supported by AWS SAM, visit AWS SAM. For more information about the Metadata resource, visit HashiCorp Terraform.

Support for the Terraform configuration is currently in preview, and the team is asking for feedback and feature request submissions. The goal is for both communities to help improve the local development process using AWS SAM CLI. Submit your feedback by creating a GitHub issue here.

For more serverless learning resources, visit Serverless Land.