14 min read

Terraform your Cloud Infrastructure

So, you have an existing infrastructure in the cloud and want to wrap it up as code in a new, shiny IaC style? Splendid! Oh… it’s spanning through two different cloud providers, has several hundred resources and services, in several different regions? Quite a pickle indeed! But don’t worry mate, I just been through a similar situation and lived to tell the tale.

Infrastructure as a code

Infrastructure as a Code definitely makes sense as one of the basic DevOps principles. Instead of manually managing cloud resources which end in usual ad-hoc solutions “because he said he needed this one EC2 instance for a moment” and creating a strange spaghetti of services “no one knows if someone uses and if can be deleted”, it forces some rules and order. Keeping infrastructure as code makes it easier to manage, review and revert changes if needed. Also, the whole infrastructure and its parts can be easily recreated in a reliable way. Terraform is also nice enough to show how the changes in code would affect the infrastructure, should the code be executed, which is an additional safety measure. Ideally of course, we should be implementing IaC from the very beginning.

Why Terraform

Terraform was created as a provisioning tool with cloud services in mind, and shines brightest when orchestrating cloud services. It supports most of the popular providers like AWS, GCP and Azure, has great documentation and is open-source and user-friendly. Puppet, Ansible, Chef and Salt are better at configuration management and Cloud Formation is AWS only.

I would assume the reader has some experience in Terraform so I won’t explain what state or resource is. I will focus on imports using AWS in examples.

Import resource

To kick things off, let’s import a single IAM user. I made a simple Terraform configuration in file main.tf:

terraform {  
  required_providers {    
    aws = {      
      source  = "hashicorp/aws"      
      version = "~> 3.27"    
  required_version = ">= 0.14.9"
provider "aws" {  
  profile = "default"  
  region  = "eu-west-1"

Next, I made a single user named “terraform_user” manually from AWS Web Console.

From there, let's see how to import an IAM user. Majority of resources supported by Terraform have an “Import” section in their documentation page.


So, in this example import takes as arguments:

type_of_resource.resource_name existing_user_name

Quick note:
Do not be misled, not every resource import is that straightforward. Some, for example aws_iam_role_policy_attachment are much more complicated:

$ terraform import aws_iam_role_policy_attachment.test-attach


Where role name and policy arn separated by /. Always check “Import” section.

With that knowledge I added a simple mock up of my user resource to my Terraform configuration:

resource "aws_iam_user" "my_user" {
    	name = "terraform_user"

Executing terraform plan shows:

  # aws_iam_user.my_user will be created
  + resource "aws_iam_user" "my_user" {
  	+ arn       	= (known after apply)
  	+ force_destroy = false
  	+ id        	= (known after apply)
  	+ name      	= "terraform_user"
  	+ path      	= "/"
  	+ tags_all  	= (known after apply)
  	+ unique_id 	= (known after apply)
Plan: 1 to add, 0 to change, 0 to destroy.

Looks ok, so let's import the user executing command:

$ terraform import aws_iam_user.my_user terraform_user

aws_iam_user.my_user: Importing from ID "terraform_user"...
aws_iam_user.my_user: Import prepared!
  Prepared aws_iam_user for import
aws_iam_user.my_user: Refreshing state... [id=terraform_user]
Import successful!
The resources that were imported are shown above. These resources are now in
your Terraform state and will henceforth be managed by Terraform.

So  the user is listed in state:

$ terraform state list


Let’s check the plan again:

$ terraform plan

Terraform will perform the following actions:
  # aws_iam_user.my_user will be updated in-place
  ~ resource "aws_iam_user" "my_user" {
  	+ force_destroy = false
    	id        	= "terraform_user"
    	name      	= "terraform_user"
  	~ tags      	= {
      	- "color" = "red" -> null
  	~ tags_all  	= {
      	- "color" = "red"
    	} -> (known after apply)
    	# (3 unchanged attributes hidden)
Plan: 0 to add, 1 to change, 0 to destroy.

The plan command found a little difference between my code and the existing user. force_destroy = false is default value so it can be ignored in this case. I will add the missing tag to my Terraform config:

resource "aws_iam_user" "my_user" {
    	name = "terraform_user"
    	tags = {
            	"color" = "red"

Now the plan shows:

No changes. Your infrastructure matches the configuration.

The job is done, the user is added to the state and managed in Terraform. Applying the plan is also safe and would not cause any changes.

In the case of mistakes or major edits, resources can be removed from state using command:

$ terraform state rm resource_type.resource_name

So in our case it will be:

$ terraform state rm aws_iam_user.my_user

Okay, now do it manually several hundred times for different resources. See ya!

...just kidding. Adding it this way may be fine for a single resource, as we needed to look up the user name in the AWS Console to import it. To efficiently import big, existing infrastructure, we need some kind of mapping tool for all kinds of resources. As usual, the open source community did not disappoint.

Meet Terraformer


Terraformer is an open source CLI tool for generating Terraform code based on existing infrastructure. It’s very good for mapping cloud resources, but won’t do all the job for you. Generated code often contains default values, which decrease readability and resource names are not that user friendly. Imports still need to be done manually by command terraform import too. Our example user has some resources attached to it, so let’s see what we get from our IAM example (region is needed in terraformer command, but it doesn’t matter for IAM service, as they are global):

$ terraformer import aws --resources="iam" --regions="eu-west-1" --profile="default"

2021/10/11 05:00:53 aws importing default region
2021/10/11 05:00:55 aws importing... iam
2021/10/11 05:00:59 aws done importing iam
2021/10/11 05:00:59 Number of resources for service iam: 9
2021/10/11 05:00:59 Refreshing state... aws_iam_role.tfer--AWSServiceRoleForSupport
2021/10/11 05:00:59 Refreshing state... aws_iam_role_policy_attachment.tfer--AWSServiceRoleForSupport_AWSSupportServiceRolePolicy
2021/10/11 05:00:59 Refreshing state... aws_iam_role_policy_attachment.tfer--AWSServiceRoleForTrustedAdvisor_AWSTrustedAdvisorServiceRolePolicy
2021/10/11 05:00:59 Refreshing state... aws_iam_user_group_membership.tfer--terraform_user-002F-Admin
2021/10/11 05:00:59 Refreshing state... aws_iam_group.tfer--Admin
2021/10/11 05:00:59 Refreshing state... aws_iam_group_policy_attachment.tfer--Admin_AmazonEC2FullAccess
2021/10/11 05:00:59 Refreshing state... aws_iam_user_policy_attachment.tfer--terraform_user_AmazonS3ReadOnlyAccess
2021/10/11 05:00:59 Refreshing state... aws_iam_user.tfer--AIDAR33EZOPUUP2XE5OJ4
2021/10/11 05:00:59 Refreshing state... aws_iam_role.tfer--AWSServiceRoleForTrustedAdvisor
2021/10/11 05:01:02 Filtered number of resources for service iam: 9
2021/10/11 05:01:02 aws Connecting....
2021/10/11 05:01:02 aws save iam
2021/10/11 05:01:02 aws save tfstate for iam

Terraformer found 9 different resources. Some of them like roles and policies are AWS managed, so we don’t have to import them. User, group, group membership and policy attachments on the other hand, are what interest us. These files are created by the terraformer:

└── generated
	└── aws
    	└── iam
        	├── iam_group_policy_attachment.tf
        	├── iam_group.tf
        	├── iam_role_policy_attachment.tf
        	├── iam_role.tf
        	├── iam_user_group_membership.tf
        	├── iam_user_policy_attachment.tf
        	├── iam_user.tf
        	├── outputs.tf
        	├── provider.tf
        	└── terraform.tfstate
3 directories, 10 files

That’s a lot for one IAM user. Let’s analyze the files. Going from the bottom:

  • terraform.tfstate is the statefile for mapped resources, can be useful in some manual statefile manipulation
  • provider.tf is a simple provider definition, very similar to my main.tf from beginning
  • outputs.tf is very crude mapping of all resources for potential reference use, containing for example:
output "aws_iam_user_tfer--AIDAR33EZOPUUP2XE5OJ4_id" {
  value = "${aws_iam_user.tfer--AIDAR33EZOPUUP2XE5OJ4.id}"

So in its current form it’s very hard to read and use, I would recommend writing your own outputs with some planning behind them.

In our scenario, those three files were not very interesting, they often can be safely ignored and deleted. The rest of them are actual mapped resources and I will omit some of them as they contain AWS managed roles and policies we don’t need to import.


resource "aws_iam_user" "tfer--AIDAR33EZOPUUP2XE5OJ4" {
  force_destroy = "false"
  name      	= "terraform_user"
  path      	= "/"
  tags = {
	color = "red"
  tags_all = {
	color = "red"

This looks almost exactly like a resource we added manually. Executing this code would of course be valid and have the exact same effect as the one we wrote, but there are still some things to consider. Attributes force_destroy, path and tags_all can be safely deleted from config, the first two are default values, and the latter is a read-only attribute only saved in state, containing all tags of resource for reference purposes. Resource name is not very informative, so it’s advised to change it too.


resource "aws_iam_user_policy_attachment" "tfer--terraform_user_AmazonS3ReadOnlyAccess" {
  policy_arn = "arn:aws:iam::aws:policy/AmazonS3ReadOnlyAccess"
  user   	= "terraform_user"


resource "aws_iam_user_group_membership" "tfer--terraform_user-002F-Admin" {
  groups = ["Admin"]
  user   = "terraform_user"


resource "aws_iam_group" "tfer--Admin" {
  name = "Admin"
  path = "/"


resource "aws_iam_group_policy_attachment" "tfer--Admin_AmazonEC2FullAccess" {
  group  	= "Admin"
  policy_arn = "arn:aws:iam::aws:policy/AmazonEC2FullAccess"

Now those files had resources we needed. It seems our terraform_user is in the Admin group which has AmazonEC2FullAccess permissions, and policy AmazonS3ReadOnlyAccess attached directly to it. I'll add those lines to my code:

resource "aws_iam_user_policy_attachment" "terraform_user_AmazonS3ReadOnlyAccess" {
  policy_arn = "arn:aws:iam::aws:policy/AmazonS3ReadOnlyAccess"
  user   	= "terraform_user"
resource "aws_iam_user_group_membership" "terraform_user-Admin" {
  groups = ["Admin"]
  user   = "terraform_user"
resource "aws_iam_group" "Admin" {
  name = "Admin"
resource "aws_iam_group_policy_attachment" "Admin_AmazonEC2FullAccess" {
  group  	= "Admin"
  policy_arn = "arn:aws:iam::aws:policy/AmazonEC2FullAccess"

I changed resource names to be more readable and deleted the path attribute from aws_iam_group as it was default and not needed.
$ terraform plan

Terraform will perform the following actions:
  # aws_iam_group.Admin will be created
  + resource "aws_iam_group" "Admin" {
  	+ arn   	= (known after apply)
  	+ id    	= (known after apply)
  	+ name  	= "Admin"
  	+ path  	= "/"
  	+ unique_id = (known after apply)
  # aws_iam_group_policy_attachment.Admin_AmazonEC2FullAccess will be created
  + resource "aws_iam_group_policy_attachment" "Admin_AmazonEC2FullAccess" {
  	+ group  	= "Admin"
  	+ id     	= (known after apply)
  	+ policy_arn = "arn:aws:iam::aws:policy/AmazonEC2FullAccess"
  # aws_iam_user_group_membership.terraform_user-Admin will be created
  + resource "aws_iam_user_group_membership" "terraform_user-Admin" {
  	+ groups = [
      	+ "Admin",
  	+ id 	= (known after apply)
  	+ user   = "terraform_user"
  # aws_iam_user_policy_attachment.terraform_user_AmazonS3ReadOnlyAccess will be created
  + resource "aws_iam_user_policy_attachment" "terraform_user_AmazonS3ReadOnlyAccess" {
  	+ id     	= (known after apply)
  	+ policy_arn = "arn:aws:iam::aws:policy/AmazonS3ReadOnlyAccess"
  	+ user   	= "terraform_user"
Plan: 4 to add, 0 to change, 0 to destroy.

Looks ok, let’s import those four resources. Writing those commands manually may be boring and cause mistakes so I advise copying strings from plan and terraformer outputs as they often contain the required ARN, or creating whole scripts via some sed/grep/vim magic. For example string form plan # aws_iam_group.Admin will be created contain the first half of our import command, grep all “will be created” replace # with “terraform import” and then just add the last part from the existing resource accordingly to the documentation.

$ terraform import aws_iam_group.Admin Admin

$ terraform import aws_iam_group_policy_attachment.Admin_AmazonEC2FullAccess

$ terraform import aws_iam_user_group_membership.terraform_user-Admin

$ terraform import aws_iam_user_policy_attachment.terraform_user_AmazonS3ReadOnlyAccess

Everything has been added correctly, and after executing the plan we get:

No changes. Your infrastructure matches the configuration.

Plus the state list shows managed resources as expected:


Things to remember using Terraformer:

  • Name is often not user friendly, sometimes can even have ARNs, it is heavily advised to change it to add readability.
  • Terraformer must be executed for specified region or regions, sometimes they may overlap so be careful.
  • Use references! Terraformer sadly won’t do it for you, so if you find in resource attributes things like ARNs, security groups or vpcs use references to make code more mobile and less prone to errors while recreating infrastructure.
  • While mapping many resources be methodical, narrow down file after file, resource after resource to get rid of AWS managed ones. There is an option to pack all resources into one file, but I found it harder to work with.

Using --compact parameter to group resource files within a single service into one resources.tf file

Importing tips and good practices:

  • If your infrastructure is spanning through several regions, remember to set the correct region for every resource. The import will fail if it doesn't find a given resource in the default region. 
  • Some resources won't have import options, for example parts of a custom module. In that case, the resource must be recreated once to be managed by terraform.
  • Do not pack everything into a single statefile as it will be hard to work with, terraform commands will take a long time to execute or even fail with timeouts. Slice state logically into big blocks like regions, environments or applications. It’s best to do it after importing, as rollback will be easier in case of mistakes.
  • When doing changes in many resources, major refactors or adding tags for example, execute theplan frequently as  to not end up with huge applies to make.
  • If you have similar resources it may make sense to enclose them in modules to minimize the codebase and make code easier to maintain. Use for_each and count for easier imports.
  • As written in Terraformer tips - user references when you can.
  • Be sure imported resources are not managed somewhere else. 
  • Discuss CI/CD procedures with your team to manage infrastructure via Terraform from this point on without creating differences between state and cloud resources.

Plan your imports, do things slowly and methodically so you don't see monsters like this:

Don't do that. Ever!

Did you like this blog post? Check out our other blogs and sign up for our newsletter to stay up to date!

big data
cloud computing
cloud infrastructure
14 October 2021

Want more? Check our articles


5 main data-related trends to be covered at Big Data Tech Warsaw 2021. Part I.

A year is definitely a long enough time to see new trends or technologies that get more traction. The Big Data landscape changes increasingly fast…

Read more
kafka gobblin hdfs getindata linkedin

Data pipeline evolution at Linkedin on a few pictures

Data Pipeline Evolution The LinkedIn Engineering blog is a great resource of technical blog posts related to building and using large-scale data…

Read more
llm data enrichment bigqueryobszar roboczy 1 4

How to use LLMs for data enrichment in BigQuery?

Introduction In the ever-evolving world of data analytics, businesses are continuously seeking innovative methods to unlock hidden value from their…

Read more
big data blog getindata data enrichment flink sql http connector

Data Enrichment in Flink SQL using HTTP Connector For Flink - Part One

HTTP Connector For Flink SQL  In our projects at GetInData, we work a lot on scaling out our client's data engineering capabilities by enabling more…

Read more

Data Journey with Michał Wróbel (RenoFi) - Doing more with less with a Modern Data Platform and ML at home

In this episode of the RadioData Podcast, Adam Kawa talks with Michał Wróbel about business use cases at RenoFi (​​a U.S.-based FinTech), the Modern…

Read more

ETL 2.0 Why you should switch into stream processing

If you are looking at Nifi to help you in your data ingestions pipeline, there might be an interesting alternative. Let’s assume we want to simply…

Read more

Contact us

Interested in our solutions?
Contact us!

Together, we will select the best Big Data solutions for your organization and build a project that will have a real impact on your organization.

What did you find most impressive about GetInData?

They did a very good job in finding people that fitted in Acast both technically as well as culturally.
Type the form or send a e-mail: hello@getindata.com
The administrator of your personal data is GetInData Poland Sp. z o.o. with its registered seat in Warsaw (02-508), 39/20 Pulawska St. Your data is processed for the purpose of provision of electronic services in accordance with the Terms & Conditions. For more information on personal data processing and your rights please see Privacy Policy.

By submitting this form, you agree to our Terms & Conditions and Privacy Policy