Explaining Count Loop in Terraform

The aim of this page📝 is to explain count looping constructs in Terraform. Being new to HCL and its declarative nature, this looping construct/meta-argument is initially weird (there is also for_each and dynamic blocks). So, you iterate, but you also assign a value (use =) simultaneously. In addition, your loop is a part of the resource definition, not the other way around.

the count loop is used to create multiple almost identical resources

• the count takes a positive integer or a 0
• why 0? because setting count to 0 is a way to condition the creation of a resource
• the definition syntax is very simple — use the keyword count right after the resource declaration
• it is possible and often useful to create a dedicated variable that is referred to in the count meta-argument instead of hardcoding the count with a number so that you control the same value used many times from a single place (classical DRY principle)

# DEFINITION 
## SYNTAX 
resource "<resource_type>" "<name_label>" {
  count = <int>
  ...
}

## EXAMPLE #1
resource "aws_instance" "web_server" {
  count = 3 
  tags = {
    Name = "mrp-web-${count.index}:
  }
}

## EXAMPLE #2
resource "aws_lb" "alb" {
  count = var.instance_count
 ...
}

# THE VARIABLE USED THROUGOUHT THE CONFIG
variable "instance count" {
  type        = number
  description = "Number of instances to create"
  default     = 2
}

there are 3 ways of referring with count: <name>[count.index] — <name>[<int>] — <name>[*]

• count creates a list of resources
• the created list is referred by classical indexing syntax using square brackets
• as terraform loops through items via count, it also creates and iterates over count.index
• count.index value starts with 0
• you explicitly use count.index for referring to resources if you are harmonizing count across resources

resource "aws_instance" "nginx" {
  count                  = var.instance_count # <-- DRY
  ami                    = nonsensitive(data.aws_ssm_parameter.amzn2_linux.value)
  instance_type          = "t2.micro"
}

# aws_lb_target_group_attachment
resource "aws_lb_target_group_attachment" "nginx" {
  count            = var.instance_count # <-- DRY
  target_group_arn = aws_lb_target_group.nginx.arn
  target_id        = aws_instance.nginx[count.index].id # <--- `count.index` refering to the first resource within larger iteration
  port             = 80
}

• what is in the language is the possibility to refer to all instances at once by using a wildcard *
• …this is a special character returning list of referred values for each instance
• using * is known as splat syntax

# REFERENCE/INDEXING 
## SYNTAX
<!--  <attribute> is optional -->

<resource_type>.<name_label>[element_index].<attribute>
<resource_type>.<name_label>[*].<attribute>

# REFERENCE EXAMPLE #1
aws_instance.web_server[count.index].name
aws_instance.web_server[0].name
aws_instance.web_server[*].name

# REFERENCE EXAMPLE #2
output "alb_id" {
  value = "${var.deploy ? join("", aws_lb.alb.*.id) : ""}"
}

LINKS

• https://developer.hashicorp.com/terraform/language/meta-arguments/count