What is the difference between count and for_each in Terraform?

Both count and for_each are Terraform meta-arguments that allow you to create multiple instances of a resource from a single resource block, but they work differently and have important trade-offs.

count

count creates resources by index (0, 1, 2, …). Resources are addressed as resource_type.name[index].

resource "aws_instance" "web" {
  count         = 3
  ami           = "ami-0c55b159cbfafe1f0"
  instance_type = "t2.micro"

  tags = {
    Name = "web-${count.index}"
  }
}

# Reference: aws_instance.web[0], aws_instance.web[1], aws_instance.web[2]

count Problem: Index-based destruction

If you delete element at index 1 from a list of 3, Terraform destroys index 1 AND recreates index 2 as the new index 1. This causes unnecessary resource churn.

for_each

for_each creates resources from a map or set of strings. Resources are addressed by their key.

# Using a set
resource "aws_iam_user" "users" {
  for_each = toset(["alice", "bob", "carol"])
  name     = each.key
}

# Using a map
resource "aws_instance" "web" {
  for_each = {
    prod = "t3.medium"
    dev  = "t3.small"
  }
  ami           = "ami-0c55b159cbfafe1f0"
  instance_type = each.value

  tags = {
    Name = "web-${each.key}"
  }
}

# Reference: aws_instance.web["prod"], aws_instance.web["dev"]

for_each Advantage: Key-based stability

Deleting “bob” from the list only destroys bob’s resource. Alice and Carol are unaffected.

Comparison

When to Use Each

Use count when:

• Creating identical resources (e.g., N identical EC2 instances)
• The total count is determined by a simple number
• Resources don’t need stable identity

Use for_each when:

• Creating resources with different configurations
• Working with a list/map of named resources
• Stability of individual resources is important (prevents unnecessary destruction)