Lab: count vs for_each for Resource Instances

Objective

By the end of this lab, you will have:

• Created a configuration with multiple instances using count.
• Created a configuration with multiple instances using for_each.
• Verified that count is identity-fragile.
• Verified that for_each is identity-stable.

Requirements

• A Linux or macOS workstation with shell access.
• The Terraform CLI 1.9.x or later installed.

Scenario

You have a configuration that needs to create multiple files. You are choosing between count and for_each. The choice has implications for refactoring and state stability.

Tasks

Task 1: Create the working directory

mkdir -p ~/rb-count-vs-foreach
cd ~/rb-count-vs-foreach
mkdir -p count foreach
cd ~/rb-count-vs-foreach

Task 2: Create the count configuration

Create count/main.tf:

terraform {
  required_version = ">= 1.9.0"
  required_providers {
    local = {
      source  = "hashicorp/local"
      version = "~> 2.5"
    }
  }
}

resource "local_file" "readme" {
  count = 3

  filename = "${path.module}/count-${count.index}.txt"
  content  = "File from count, index ${count.index}\n"
}

Task 3: Apply the count configuration

cd count
terraform init
terraform apply

Verify:

ls count-*.txt
cat count-0.txt

Task 4: Modify the count

Change the count from 3 to 2:

resource "local_file" "readme" {
  count = 2
  ...
}

Apply:

terraform apply

Verify:

ls count-*.txt

The count-2.txt file is destroyed. The remaining files are count-0.txt and count-1.txt.

This is the count antipattern: changing the count can destroy real-world resources.

Task 5: Create the for_each configuration

Create foreach/main.tf:

terraform {
  required_version = ">= 1.9.0"
  required_providers {
    local = {
      source  = "hashicorp/local"
      version = "~> 2.5"
    }
  }
}

resource "local_file" "readme" {
  for_each = toset(["a", "b", "c"])

  filename = "${path.module}/foreach-${each.key}.txt"
  content  = "File from for_each, key ${each.key}\n"
}

Task 6: Apply the for_each configuration

cd ../foreach
terraform init
terraform apply

Verify:

ls foreach-*.txt
cat foreach-a.txt

Task 7: Modify the for_each

Remove the key “b” from the set:

resource "local_file" "readme" {
  for_each = toset(["a", "c"])
  ...
}

Apply:

terraform apply

Verify:

ls foreach-*.txt

The foreach-b.txt file is destroyed. The foreach-a.txt and foreach-c.txt files are unchanged.

This is the for_each advantage: changing the keys only affects the changed keys.

Task 8: Add a new key

Add the key “d” to the set:

resource "local_file" "readme" {
  for_each = toset(["a", "c", "d"])
  ...
}

Apply:

terraform apply

Verify:

ls foreach-*.txt

The foreach-d.txt file is created.

Validation

The lab is successful if:

• The count configuration created 3 files and then destroyed 1.
• The for_each configuration created 3 files, destroyed 1, and created 1.

Expected Outcome

After the count experiment:

count-0.txt
count-1.txt

After the for_each experiment:

foreach-a.txt
foreach-c.txt
foreach-d.txt

Cleanup

cd ~/rb-count-vs-foreach
rm -rf count foreach

What You Learned

You learned the count vs for_each trade-off:

• count is identity-fragile. Changing the count can destroy real-world resources.
• for_each is identity-stable. Changing the keys only affects the changed keys.
• For stateful resources, use for_each with a stable semantic key.
• For ephemeral resources, count is fine.