Skip to content
About Me
----

Meta-Arguments & Resource Lifecycle Control

Architect Key Takeaways

Master Terraform meta-arguments: count, for_each, depends_on, provider alias, and lifecycle blocks (create_before_destroy, prevent_destroy, ignore_changes).

Meta-Arguments & Resource Lifecycle Control

Section titled “Meta-Arguments & Resource Lifecycle Control”

Meta-arguments change how resources are evaluated, provisioned, and destroyed by Terraform’s Directed Acyclic Graph (DAG) engine.


1. count vs. for_each: The Definitive Guide

Section titled “1. count vs. for_each: The Definitive Guide”

Using count indexes resources by array integers ([0], [1], [2]). If an item in the middle of the list is deleted, all subsequent resources shift index, forcing Terraform to destroy and recreate live infrastructure!

# ❌ DANGEROUS: If "marketing" is removed, "finance" becomes index 1 and gets recreated!
variable "departments" {
default = ["engineering", "marketing", "finance"]
}
resource "aws_iam_user" "dept" {
count = length(var.departments)
name = var.departments[count.index]
}

for_each binds each resource to a persistent string key, guaranteeing that adding or deleting keys never affects adjacent resources:

# ✅ PRODUCTION SAFE: Each user is identified by unique key ("engineering", "marketing", etc.)
resource "aws_iam_user" "dept" {
for_each = toset(["engineering", "marketing", "finance"])
name = each.key
}

Terraform automatically builds resource dependency trees from referenced attributes (e.g. subnet_id = aws_subnet.pub.id). However, when dependencies are indirect (such as an IAM role needing time to propagate before an EKS cluster or Lambda function uses it), use depends_on:

resource "aws_eks_cluster" "main" {
name = "enterprise-eks-cluster"
role_arn = aws_iam_role.cluster.arn
# Explicitly wait until AWS IAM policy attachment completes
depends_on = [
aws_iam_role_policy_attachment.cluster_AmazonEKSClusterPolicy,
aws_iam_role_policy_attachment.cluster_AmazonEKSVPCResourceController
]
}

3. The lifecycle Block: Controlling Provisioning Behavior

Section titled “3. The lifecycle Block: Controlling Provisioning Behavior”

The lifecycle meta-argument allows fine-grained control over how changes are applied:

resource "aws_instance" "app_node" {
ami = var.ami_id
instance_type = "t3.large"
lifecycle {
# 1. Zero-Downtime Rollout: Creates new node before terminating old one
create_before_destroy = true
# 2. Safety Net: Prevents accidental deletion of mission-critical assets
prevent_destroy = false
# 3. Drift Suppression: Ignores changes made by external autoscalers or security scanners
ignore_changes = [
tags["LastSecurityAudit"],
user_data
]
}
}

4. Multi-Region & Multi-Account with provider Alias

Section titled “4. Multi-Region & Multi-Account with provider Alias”

When provisioning cross-region resources (e.g., CloudFront ACM certificates in us-east-1 while hosting infrastructure in eu-west-1):

provider "aws" {
alias = "us_east_1"
region = "us-east-1"
}
resource "aws_acm_certificate" "cdn_cert" {
provider = aws.us_east_1 # Directs certificate creation to us-east-1
domain_name = "devsecopsguru.in"
validation_method = "DNS"
}