Terraform Import

Terraform can import existing infrastructure, which means you can import AWS resources created by some other method to be henceforth managed by Terraform and your IaC.

The terraform import command modifies the state and must be executed by an AWS Account Admin.

1. (Developer) Write a resource block and push the code changes:

# Example
resource "aws_cloudwatch_log_group" "example" {
  name              = "/aws/lambda/${var.lambda_function_name}"
  retention_in_days = 14
}

2. (Admin) Login to the AWS account with the aws CLI
3. (Admin) terraform import doesn't work with a dynamic provider config (known limitation).
   • Temporarily/locally hard-code the region in the provider block:

   # DOES NOT WORK with import command:
   provider "aws" {
       region = var.region
   }
   
   # WORKS with import command:
   provider "aws" {
       region = "us-east-2"
   }
   

4. (Admin) Locally, in the iac/{env} directory where the state file is, run:
   1. terraform init
   2. terraform import <resource address in your IaC> <existing id in AWS>

   # Example
   terraform import module.build_lambda.aws_cloudwatch_log_group.example /aws/lambda/example-fxn
   

5. (Developer) Build/deploy.
   • Now that the existing resource is recorded in your state, terraform apply shouldn't yield a ResourceAlreadyExistsException.
6. (Admin) Discard the change to hard-code the provider region.

Example Use Case: Expire Lambda Logs

Scenario

AWS Lambda functions automatically store log output in Amazon CloudWatch Logs by creating a CloudWatch Logs group for your Lambda function, named /aws/lambda/{lambda-function-name}.

Optionally, you could explicitly manage the CloudWatch Log group for a Lambda in your IaC.

resource "aws_lambda_function" "datahub_lambda" {
    function_name    = var.lambda_function_name
    ...
}

# optional 
resource "aws_cloudwatch_log_group" "example" {
  name              = "/aws/lambda/${var.lambda_function_name}"
  retention_in_days = 14
}

Managing the Log Group in your IaC could be beneficial in order to set a log retention (logs that never expire accumulate and become difficult to search).

Problem

The log group you create via Terraform for your Lambda function must be named /aws/lambda/{lambda-function-name}.

If the Lambda function is already live and has already created its CloudWatch Log Group with this name, when you add the Log Group resource to your IaC, terraform apply will fail:

Error: Creating CloudWatch Log Group failed: 
ResourceAlreadyExistsException: The specified log group already exists.

Solution

Import the existing Log Group to attach it to your IaC. This records the existing resource in your Terraform state file so it can be henceforth managed by Terraform.