devops-exercises/topics/terraform/README.md

Terraform

• Terraform
  • Exercises
    • Terraform 101
    • AWS
  • Questions
    • Terraform 101
    • Terraform Hands-On Basics
    • Dependencies
    • Providers
    • Variables
      • Input Variables
      • Output Variables
      • Variables Hands-On
    • Data Sources
    • Lifecycle
    • Provisioners
    • State
      • Terraform Backend
      • Workspaces
      • State Hands-On
    • Modules
    • Import
    • Version Control
    • AWS
    • Validations
    • Terraform Syntax

Exercises

Terraform 101

Name	Topic	Objective & Instructions	Solution	Comments
Local Provider	Basics	Exercise	Solution

AWS

The following exercises require account in AWS and might cost you $

Name	Topic	Objective & Instructions	Solution	Comments
Launch EC2 instance	EC2	Exercise	Solution
Rename S3 bucket	S3	Exercise	Solution

Questions

Terraform 101

What is Terraform?

Terraform: "HashiCorp Terraform is an infrastructure as code tool that lets you define both cloud and on-prem resources in human-readable configuration files that you can version, reuse, and share. You can then use a consistent workflow to provision and manage all of your infrastructure throughout its lifecycle. Terraform can manage low-level components like compute, storage, and networking resources, as well as high-level components like DNS entries and SaaS features."

What are the advantages in using Terraform or IaC in general?

• Full automation: In the past, resource creation, modification and removal were handled manually or by using a set of tooling. With Terraform or other IaC technologies, you manage the full lifecycle in an automated fashion.
  
• Modular and Reusable: Code that you write for certain purposes can be used and assembled in different ways. You can write code to create resources on a public cloud and it can be shared with other teams who can also use it in their account on the same (or different) cloud>
  
• Improved testing: Concepts like CI can be easily applied on IaC based projects and code snippets. This allow you to test and verify operations beforehand

What are some of Terraform features?

• Declarative: Terraform uses the declarative approach (rather than the procedural one) in order to define end-status of the resources
• No agents: as opposed to other technologies (e.g. Puppet) where you use a model of agent and server, with Terraform you use the different APIs (of clouds, services, etc.) to perform the operations
• Community: Terraform has strong community who constantly publishes modules and fixes when needed. This ensures there is good modules maintenance and users can get support quite quickly at any point

What language does Terraform uses?

A DSL called "HCL" (Hashiciorp Configuration Language). A declarative language for defining infrastructure.

What's a typical Terraform workflow?

1. Write Terraform definitions: .tf files written in HCL that described the desired infrastructure state (and run terraform init at the very beginning)
2. Review: With command such as terraform plan you can get a glance at what Terraform will perform with the written definitions
3. Apply definitions: With the command terraform apply Terraform will apply the given definitions, by adding, modifying or removing the resources

This is a manual process. Most of the time this is automated so user submits a PR/MR to propose terraform changes, there is a process to test these changes and once merged they are applied (terraform apply).

What are some use cases for using Terraform?

• Infra provisioning and management: You need to automated or code your infra so you are able to test it easily, apply it and make any changes necessary.
• Multi-cloud environment: You manage infrastructure on different clouds, but looking for a consistent way to do it across the clouds
• Consistent environments: You manage environments such as test, production, staging, ... and looking for a way to have them consistent so any modification in one of them, applies to other environments as well

What's the difference between Terraform and technologies such as Ansible, Puppet, Chef, etc.

Terraform is considered to be an IaC technology. It's used for provisioning resources, for managing infrastructure on different platforms.

Ansible, Puppet and Chef are Configuration Management technologies. They are used once there is an instance running and you would like to apply some configuration on it like installing an application, applying security policy, etc.

To be clear, CM tools can be used to provision resources so in the end goal of having infrastructure, both Terraform and something like Ansible, can achieve the same result. The difference is in the how. Ansible doesn't saves the state of resources, it doesn't know how many instances there are in your environment as opposed to Terraform. At the same time while Terraform can perform configuration management tasks, it has less modules support for that specific goal and it doesn't track the task execution state as Ansible. The differences are there and it's most of the time recommended to mix the technologies, so Terraform used for managing infrastructure and CM technologies used for configuration on top of that infrastructure.

Terraform Hands-On Basics

Explain the following block of Terraform code

resource "aws_instance" "some-instance" {
  ami           = "ami-201720221991yay"
  instance_type = "t2.micro
}

It's a resource of type "aws_instance" used to provision an instance. The name of the resource (NOT INSTANCE) is "some-instance".

The instance itself will be provisioned with type "t2.micro" and using an image of the AMI "ami-201720221991yay".

What do you do next after writing the following in main.tf file?

resource "aws_instance" "some-instance" {
  ami           = "ami-201720221991yay"
  instance_type = "t2.micro
}

Run terraform init. This will scan the code in the directory to figure out which providers are used (in this case AWS provider) and will download them.

You've executed terraform init and now you would like to move forward to creating the resources but you have concerns and would like to make be 100% sure on what you are going to execute. What should you be doing?

Execute terraform plan. That will provide a detailed information on what Terraform will do once you apply the changes.

You've downloaded the providers, seen the what Terraform will do (with terraform plan) and you are ready to actually apply the changes. What should you do next?

Run terraform apply. That will apply the changes described in your .tf files.

Explain the meaning of the following strings that seen at the beginning of each line When you run terraform apply

• '+'
• '-'
• '-/+'

• '+' - The resource or attribute is going to be added
• '-' - the resource or attribute is going to be removed
• '-/+' - the resource or attribute is going to be replaced

Dependencies

Sometimes you need to reference some resources in the same or separate .tf file. Why and how it's done?

Why: because resources are sometimes connected or need to be connected. For example, you create an AWS instance with "aws_instance" resource but, at the same time you would like also to allow some traffic to it (because by default traffic is not allowed). For that you'll create a "aws_security_group" resource and then, in your aws_instance resource, you'll reference it.

How:

Using the syntax ..

In your AWS instance it would like that:

resource "aws_instance" "some-instance" {
  
  ami           = "some-ami"
  instance_type = "t2.micro"
  vpc_security_group_ids = [aws_security_group.instance.id]

}

Does it matter in which order Terraform creates resources?

Yes, when there is a dependency between different Terraform resources, you want the resources to be created in the right order and this is exactly what Terraform does.

To make it ever more clear, if you have a resource X that references the ID of resource Y, it doesn't makes sense to create first resource X because it won't have any ID to get from a resource that wasn't created yet.

Is there a way to print/see the dependencies between the different resources?

Yes, with terraform graph

The output is in DOT - A graph description language.

Providers

Explain what is a "provider"

terraform.io: "Terraform relies on plugins called "providers" to interact with cloud providers, SaaS providers, and other APIs...Each provider adds a set of resource types and/or data sources that Terraform can manage. Every resource type is implemented by a provider; without providers, Terraform can't manage any kind of infrastructure."

Where can you find publicly available providers?

In the Terraform Registry

What are the names of the providers in this case?

terraform {
    required_providers {
      aws = {
        source  = "hashicorp/aws"
      }
      azurerm = {
        source  = "hashicorp/azurerm"
        version = "~> 3.0.2"
      }
    }
  }

azurerm and aws

True or False? Applying the following Terraform configuration will fail since no source or version specific for 'aws' provider

terraform {
    required_providers {
      aws = {}
    }
  }

False. It will look for "aws" provider in the public Terraform registry and will take the latest version.

Write a configuration of a Terraform provider (any type you would like)

AWS is one of the most popular providers in Terraform. Here is an example of how to configure it to use one specific region and specifying a specific version of the provider

terraform {
  required_providers {
    aws = {
      source  = "hashicorp/aws"
      version = "~> 3.0"
    }
  }
}

# Configure the AWS Provider
provider "aws" {
  region = "us-west-2"
}

Where Terraform installs providers from by default?

By default Terraform providers are installed from Terraform Registry

What is the Terraform Registry?

The Terraform Registry provides a centralized location for official and community-managed providers and modules.

Where providers are downloaded to? (when for example you run terraform init)

.terraform directory.

How to cleanup Terraform resourcse? Why the user shold be careful doing so?

terraform destroy will cleanup all the resources tracked by Terraform.

A user should be careful with this command because there is no way to revert it. Sure, you can always run again "apply" but that can take time, generates completely new resources, etc.

Variables

Input Variables

What input variables are good for in Terraform?

Variables allow you define piece of data in one location instead of repeating the hardcoded value of it in multiple different locations. Then when you need to modify the variable's value, you do it in one location instead of changing each one of the hardcoded values.

What type of input variables are supported in Terraform?

string
number
bool
list(<TYPE>)
set(<TYPE>)
map(<TYPE>)
object({<ATTR_NAME> = <TYPE>, ... })
tuple([<TYPE>, ...])

What's the default input variable type in Terraform?

any

What ways are there to pass values for input variables?

• Using -var option in the CLI
• Using a file by using the -var-file option in the CLI
• Environment variable that starts with TF_VAR_<VAR_NAME>

If no value given, user will be prompted to provide one.

How to reference a variable?

Using the syntax var.<VAR_NAME>

What is the effect of setting variable as "sensitive"?

It doesn't show its value when you run terraform apply or terraform plan but eventually it's still recorded in the state file.

True or False? If an expression's result depends on a sensitive variable, it will be treated as sensitive as well

True

The same variable is defined in the following places:

• The file terraform.tfvars
• Environment variable
• Using -var or -var-file

According to variable precedence, which source will be used first?

The order is:

• Environment variable
• The file terraform.tfvars
• Using -var or -var-file

Whenever you run terraform apply, it prompts to enter a value for a given variable. How to avoid being prompted?

While removing the variable is theoretically a correct answer, it will probably fail the execution.

You can use something like the -var option to provide the value and avoid being prompted to insert a value. Another option is to run export TF_VAR_<VAR_NAME>=<VALUE>.

Output Variables

What are output variables? Why do we need them?

Output variable allow you to display/print certain piece of data as part of Terraform execution.

The most common use case for it is probably to print the IP address of an instance. Imagine you provision an instance and you would like to know what the IP address to connect to it. Instead of looking for it for the console/OS, you can use the output variable and print that piece of information to the screen

Explain the "sensitive" parameter of output variable

When set to "true", Terraform will avoid logging output variable's data. The use case for it is sensitive data such as password or private keys.

Explain the "depends" parameter of output variable

It is used to set explicitly dependency between the output variable and any other resource. Use case: some piece of information is available only once another resource is ready.

Variables Hands-On

Demonstrate input variable definition with type, description and default parameters

variable "app_id" {
  type = string
  description = "The id of application"
  default = "some_value"
}

Unrelated note: variables are usually defined in their own file (vars.tf for example).

How to define an input variable which is a list of numbers?

variable "list_of_nums" {
  type = list(number)
  description = "An example of list of numbers"
  default = [2, 0, 1, 7]
}

How to define an input variable which is an object with attributes "model" (string), "color" (string), year (number)?

variable "car_model" {
  description = "Car model object"
  type        = object({
    model   = string
    color   = string
    year    = number
  })
}

Note: you can also define a default for it.

How to reference variables?

Variable are referenced with var.VARIABLE_NAME syntax. Let's have a look at an example:

vars.tf:

variable "memory" {
  type = string
  default "8192"
}

variable "cpu" {
  type = string
  default = "4"
}

main.tf:

resource "libvirt_domain" "vm1" {
   name = "vm1"
   memory = var.memory
   cpu = var.cpu
}

How to reference variable from inside of string literal? (bonus question: how that type of expression is called?)

Using the syntax: "${var.VAR_NAME}". It's called "interpolation".

Very common to see it used in user_data attribute related to instances.

user_data = <<-EOF
            This is some fabulos string
            It demonstrates how to use interpolation
            Yes, it's truly ${var.awesome_or_meh}
            EOF

How can list all outputs without applying Terraform changes?

terraform output will list all outputs without applying any changes

Can you see the output of specific variable without applying terrafom changes?

Yes, with terraform output <OUTPUT_VAR>.

Very useful for scripts :)

Data Sources

Explain data sources in Terraform

• Data sources used to get data from providers or in general from external resources to Terraform (e.g. public clouds like AWS, GCP, Azure).
• Data sources used for reading. They are not modifying or creating anything
• Many providers expose multiple data sources

Demonstrate how to use data sources

data "aws_vpc" "default {
  default = true
}

How to get data out of a data source?

The general syntax is data.<PROVIDER_AND_TYPE>.<NAME>.<ATTRBIUTE>

So if you defined the following data source

data "aws_vpc" "default {
  default = true
}

You can retrieve the ID attribute this way: data.aws_vpc.default.id

Is there such a thing as combining data sources? What would be the use case?

Yes, you can define a data source while using another data source as a filter for example.

Let's say we want to get AWS subnets but only from our default VPC:

data "aws_subnets" "default" {
  filter {
    name   = "vpc-id"
    values = [data.aws_vpc.default.id]
  }
}

Lifecycle

When you update a resource, how it works?

By default the current resource is deleted, a new one is created and any references pointing the old resource are updated to point the new resource

Is it possible to modify the default lifecycle? How? Why?

Yes, it's possible. There are different lifecycles one can choose from. For example "create_before_destroy" which inverts the order and first creates the new resource, updates all the references from old resource to the new resource and then removes the old resource.

How to use it:

lifecycle {
  create_before_destroy = true
}

Why to use it in the first place: you might have resources that have dependency where they dependency itself is immutable (= you can't modify it hence you have to create a new one), in such case the default lifecycle won't work because you won't be able to remove the resource that has the dependency as it still references an old resource. AWS ASG + launch configurations is a good example of such use case.

You've deployed a virtual machine with Terraform and you would like to pass data to it (or execute some commands). Which concept of Terraform would you use?

Provisioners

Provisioners

What are "Provisioners"? What they are used for?

Provisioners can be described as plugin to use with Terraform, usually focusing on the aspect of service configuration and make it operational.

Few example of provisioners:

• Run configuration management on a provisioned instance using technology like Ansible, Chef or Puppet.
• Copying files
• Executing remote scripts

Why is it often recommended to use provisioners as last resort?

Since a provisioner can run a variety of actions, it's not always feasible to plan and understand what will happen when running a certain provisioner. For this reason, it's usually recommended to use Terraform built-in option, whenever's possible.

What is local-exec and remote-exec in the context of provisioners?

What is a "tainted resource"?

It's a resource which was successfully created but failed during provisioning. Terraform will fail and mark this resource as "tainted".

What terraform taint does?

terraform taint resource.id manually marks the resource as tainted in the state file. So when you run terraform apply the next time, the resource will be destroyed and recreated.

What is a data source? In what scenarios for example would need to use it?

Data sources lookup or compute values that can be used elsewhere in terraform configuration.

There are quite a few cases you might need to use them:

• you want to reference resources not managed through terraform
• you want to reference resources managed by a different terraform module
• you want to cleanly compute a value with typechecking, such as with aws_iam_policy_document

What are output variables and what terraform output does?

Output variables are named values that are sourced from the attributes of a module. They are stored in terraform state, and can be used by other modules through remote_state

Explain remote-exec and local-exec

Explain "Remote State". When would you use it and how?

Terraform generates a `terraform.tfstate` json file that describes components/service provisioned on the specified provider. Remote State stores this file in a remote storage media to enable collaboration amongst team.

Explain "State Locking"

State locking is a mechanism that blocks an operations against a specific state file from multiple callers so as to avoid conflicting operations from different team members. Once the first caller's operation's lock is released the other team member may go ahead to carryout his own operation. Nevertheless Terraform will first check the state file to see if the desired resource already exist and if not it goes ahead to create it.

Aside from .tfvars files or CLI arguments, how can you inject dependencies from other modules?

The built-in terraform way would be to use remote-state to lookup the outputs from other modules. It is also common in the community to use a tool called terragrunt to explicitly inject variables between modules.

How do you import existing resource using Terraform import?

1. Identify which resource you want to import.
2. Write terraform code matching configuration of that resource.
3. Run terraform command terraform import RESOURCE ID
   

eg. Let's say you want to import an aws instance. Then you'll perform following:

1. Identify that aws instance in console
2. Refer to it's configuration and write Terraform code which will look something like:

resource "aws_instance" "tf_aws_instance" {
  ami           = data.aws_ami.ubuntu.id
  instance_type = "t3.micro"

  tags = {
    Name = "import-me"
  }
}

3. Run terraform command terraform import aws_instance.tf_aws_instance i-12345678

State

What's Terraform State?

terraform.io: "Terraform must store state about your managed infrastructure and configuration. This state is used by Terraform to map real world resources to your configuration, keep track of metadata, and to improve performance for large infrastructures."

In other words, it's a mechanism in Terraform to track resources you've created or cleaned up. This is how terraform knows what to update/create/delete when you run terraform apply and also other commands like terraform destroy.

Where Terraform state is stored?

There is more than one answer to this question. It's very much depends on whether you share it with others or it's only local in your Terraform directory, but taking a beginner's case, when you run terraform in a directory, the state will be stored in that directory in terraform.tfstate file.

Can you name three different things included in the state file?

• The representation of resources - JSON format of the resources, their attributes, IDs, ... everything that required to identify the resource and also anything that was included in the .tf files on these resources
• Terraform version
• Outputs

Why does it matter where you store the tfstate file? In your answer make sure to address the following:

• Public vs. Private
• Git repository vs. Other locations

• tfstate contains credentials in plain text. You don't want to put it in publicly shared location
• tfstate shouldn't be modified concurrently so putting it in a shared location available for everyone with "write" permissions might lead to issues. (Terraform remote state doesn't has this problem).
• tfstate is an important file. As such, it might be better to put it in a location that has regular backups and good security.

As such, tfstate shouldn't be stored in git repositories. secured storage such as secured buckets, is a better option.

True or False? it's common to edit terraform state file directly by hand and even recommended for many different use cases

False. You should avoid as much possible to edit Terraform state files directly by hand.

Why storing state file locally on your computer may be problematic?

In general, storing state file on your computer isn't a problem. It starts to be a problem when you are part of a team that uses Terraform and then you would like to make sure it's shared. In addition to being shared, you want to be able to handle the fact that different teams members can edit the file and can do it at the same time, so locking is quite an important aspect as well.

Mention some best practices related to tfstate

• Don't edit it manually. tfstate was designed to be manipulated by terraform and not by users directly.
• Store it in secured location (since it can include credentials and sensitive data in general)
• Backup it regularly so you can roll-back easily when needed
• Store it in remote shared storage. This is especially needed when working in a team and the state can be updated by any of the team members
• Enabled versioning if the storage where you store the state file, supports it. Versioning is great for backups and roll-backs in case of an issue.

How and why concurrent edits of the state file should be avoided?

If there are two users or processes concurrently editing the state file it can result in invalid state file that doesn't actually represents the state of resources.

To avoid that, Terraform can apply state locking if the backend supports that. For example, AWS s3 supports state locking and consistency via DynamoDB. Often, if the backend supports it, Terraform will make use of state locking automatically so nothing is required from the user to activate it.

Describe how you manage state file(s) when you have multiple environments (e.g. development, staging and production)

Probably no right or wrong answer here, but it seems, based on different source, that the overall preferred way is to have a dedicated state file per environment.

Why storing the state in versioned control repo is not a good idea?

• Sensitive data: some resources may specify sensitive data (like passwords and tokens) and everything in a state file is stored in plain text
• Prone to errors: when working with Git repos, you mayo often find yourself switch branches, checkout specific commits, perform rebases, ... all these operations may end up in you eventually performing terraform apply on non-latest version of your Terraform code

Terraform Backend

What's a Terraform backend? What is the default backend?

Terraform backend determines how the Terraform state is stored and loaded. By default the state is local, but it's possible to set a remote backend

Describe how to set a remote backend of any type you choose

Let's say we chose use Amazon s3 as a remote Terraform backend where we can store Terraform's state.

1. Write Terraform code for creating an s3 bucket
   1. It would be a good idea to add lifecycle of "prevent_destroy" to it so it's not accidentally deleted
2. Enable versioning (add a resource of "aws_s3_bucket_versioning")
3. Encrypt the bucket ("aws_s3_bucket_server_side_encryption_configuration")
4. Block public access
5. Handle locking. One way is to add DB for it
6. Add the point you'll want to run init and apply commands to avoid an issue where you at the same time create the resources for remote backend and also switch to a remote backend
7. Once resources were created, add Terraform backend code

terraform {
  backend "s3" {
    bucket ...
  }
}

7. Run teraform init as it will configure the backend

How terraform apply workflow is different when a remote backend is used?

It starts with acquiring a state lock so others can't modify the state at the same time.

What would be te process of switching back from remote backend to local?

1. You remove the backend code and perform terraform init to switch back to local backend
2. You remove the resources that are the remote backend itself

True or False? it's NOT possible to use variable in a backend configuration

That's true and quite a limitation as it means you'll have to go to the resources of the remote backend and copy some values to the backend configuration.

One way to deal with it is using partial configurations in a completel separate file from the backend itself and then load them with terraform init -backend-config=some_backend_partial_conf.hcl

Workspaces

Explain what is a Terraform workspace

developer.hashicorp.com: "The persistent data stored in the backend belongs to a workspace. The backend initially has only one workspace containing one Terraform state associated with that configuration. Some backends support multiple named workspaces, allowing multiple states to be associated with a single configuration."

True or False? Each workspace has its own state file

True

State Hands-On

Which command will produce a state file?

terraform apply

How to inspect current state?

terraform show

How to list resources created with Terraform?

terraform state list

How do you rename an existing resource?

terraform state mv

How to create a new workspace?

terraform workspace new <WORKSPACE_NAME>

How to identify which workspace are you using?

terraform workspace show

Modules

Explain Modules

Terraform.io: "A module is a container for multiple resources that are used together. Modules can be used to create lightweight abstractions, so that you can describe your infrastructure in terms of its architecture, rather than directly in terms of physical objects."

How do you test a terraform module?

There are multiple answers, but the most common answer would likely to be using the tool terratest, and to test that a module can be initialized, can create resources, and can destroy those resources cleanly.

Where can you obtain Terraform modules?

Terraform modules can be found at the Terrafrom registry

There's a discussion in your team whether to store modules in one centralized location/repository or have them in each of the projects/repositories where they are used. What's your take on that?

You might have a different opinion but my personal take on that, is to keep modules in one centralized repository as any maintenance or updates to the module you need to perform, are done in one place instead of multiple times in different repositories.

Import

Explain Terraform's import functionality

terraform import is a CLI command used for importing an existing infrastructure into Terraform's state.

It's does NOT create the definitions/configuration for creating such infrastructure

State two use cases where you would use terraform import

1. You have existing resources in the cloud and they are not managed by Terraform (as in not included in the state)
2. You lost your tfstate file and need to rebuild it

Version Control

You have a Git repository with Terraform files but no .gitignore. What would you add to a .gitignore file in Terraform repository?

.terraform
*.tfstate
*.tfstate.backup

You don't want to store state file nor any downloaded providers in .terraform directory. It also doesn't makes sense to share/store the state backup files.

AWS

What happens if you update user_data in the following case apply the changes?

resource "aws_instance" "example" {
 ami = "..."
 instance_type = "t2.micro"

user_data = <<-EOF
            #!/bin/bash
            echo "Hello, World" > index.xhtml
            EOF

Nothing, because user_data is executed on boot so if an instance is already running, it won't change anything.

To make it effective you'll have to use user_data_replace_on_change = true.

You manage ASG with Terraform which means you also have "aws_launch_configuration" resources. The problem is that launch configurations are immutable and sometimes you need to change them. This creates a problem where Terraform isn't able to delete ASG because they reference old launch configuration. How to do deal with it?

Add the following to "aws_launch_configuration" resource

lifecycle {
  create_before_destroy = true
}

This will change the order of how Terraform works. First it will create the new resource (launch configuration). then it will update other resources to reference the new launch configuration and finally, it will remove old resources

Validations

How would you enforce users that use your variables to provide values with certain constraints? For example, a number greater than 1

Using validation block

variable "some_var" {
  type = number

  validation {
    condition = var.some_var > 1
    error_message = "you have to specify a number greater than 1"
  }

}

Terraform Syntax

Demonstrate using the ternary syntax