Ephemeral Cloud Clusters#

Ephemeral clusters exist for one purpose: validate something, then disappear. They are not staging environments, not shared dev clusters, not long-lived resources that someone forgets to turn off. The operational model is strict – create, validate, destroy – and the entire sequence must be automated so that destruction cannot be forgotten.

The cost of getting this wrong is real. A three-node EKS cluster left running over a weekend costs roughly $15. Left running for a month, $200. Multiply by the number of developers or CI pipelines that create clusters, and forgotten ephemeral infrastructure becomes a significant budget line item. Every template in this article includes auto-destroy mechanisms to prevent this.

Infrastructure Disaster Recovery with Terraform#

Application disaster recovery is well-understood: replicate data, failover traffic, restore from backups. Infrastructure disaster recovery is different — you are recovering the platform that applications run on. If your Terraform state is lost, your VPC is deleted, or an entire region goes down, how do you rebuild?

This article covers the DR patterns specific to Terraform-managed infrastructure: protecting state, recovering from state loss, designing infrastructure for regional failover, and the runbooks that agents and operators need when things go wrong.

Multi-Account Cloud Architecture with Terraform#

Single-account cloud deployments work for learning and prototypes. Production systems need multiple accounts (AWS), subscriptions (Azure), or projects (GCP) for isolation — security boundaries, blast radius control, billing separation, and compliance requirements.

Terraform manages multi-account architectures well, but the patterns differ significantly from single-account work. Provider configuration, state isolation, cross-account references, and IAM trust relationships all need explicit design.

Why Multiple Accounts#

AWS Organizations#

Organization Structure#

Organization Root
├── Core OU
│   ├── Management Account (billing, org management)
│   ├── Security Account (GuardDuty, SecurityHub, audit logs)
│   └── Networking Account (Transit Gateway, shared VPCs)
├── Workload OU
│   ├── Production OU
│   │   ├── App-A Production Account
│   │   └── App-B Production Account
│   └── Non-Production OU
│       ├── App-A Development Account
│       └── App-A Staging Account
└── Sandbox OU
    └── Developer Sandbox Accounts

Terraform for AWS Organizations#

resource "aws_organizations_organization" "main" {
  feature_set = "ALL"

  enabled_policy_types = [
    "SERVICE_CONTROL_POLICY",
    "TAG_POLICY",
  ]
}

resource "aws_organizations_organizational_unit" "core" {
  name      = "Core"
  parent_id = aws_organizations_organization.main.roots[0].id
}

resource "aws_organizations_organizational_unit" "workloads" {
  name      = "Workloads"
  parent_id = aws_organizations_organization.main.roots[0].id
}

resource "aws_organizations_organizational_unit" "production" {
  name      = "Production"
  parent_id = aws_organizations_organizational_unit.workloads.id
}

# Create a workload account
resource "aws_organizations_account" "app_production" {
  name      = "app-a-production"
  email     = "aws+app-a-prod@example.com"
  parent_id = aws_organizations_organizational_unit.production.id
  role_name = "OrganizationAccountAccessRole"  # cross-account admin role

  lifecycle {
    prevent_destroy = true  # accounts cannot be easily recreated
  }
}

Service Control Policies (SCPs)#

SCPs set permission boundaries for entire OUs:

Multi-Cloud Networking Patterns#

Multi-cloud networking connects workloads across two or more cloud providers into a coherent network. The motivations vary – vendor redundancy, best-of-breed service selection, regulatory requirements – but the challenges are the same: private connectivity between isolated networks, consistent service discovery, and traffic routing that handles failures.

VPN Tunnels Between Clouds#

IPsec VPN tunnels are the simplest way to connect two cloud networks. Each provider offers managed VPN gateways that terminate IPsec tunnels, encrypting traffic between VPCs without exposing it to the public internet.

Running Windows Workloads on Kubernetes#

Kubernetes supports Windows worker nodes alongside Linux worker nodes in the same cluster. This enables running Windows-native applications – .NET Framework services, IIS-hosted applications, Windows-specific middleware – on Kubernetes without rewriting them for Linux. However, Windows nodes are not interchangeable with Linux nodes. There are fundamental differences in networking, storage, container runtime behavior, and resource management that you must account for.

Core Constraints#

Before adding Windows nodes, understand what is and is not supported:

Service Account Security#

Every pod in Kubernetes runs as a service account. By default, that is the default service account in the pod’s namespace, with an auto-mounted API token that never expires. This default configuration is overly permissive for most workloads. Hardening service accounts is one of the highest-impact security improvements you can make in a Kubernetes cluster.

The Default Problem#

When a pod starts without specifying a service account, Kubernetes does three things:

Terraform Cloud Architecture Patterns#

The three-tier architecture — networking, managed Kubernetes, managed database — is the most common pattern for production deployments on any major cloud. The concepts are identical across AWS, Azure, and GCP. The Terraform code is not. Resource names differ, required arguments differ, default behaviors differ, and the gotchas that catch agents and humans are cloud-specific.

This article shows the real Terraform for each layer on each cloud, side by side, so agents can write correct infrastructure code for whichever cloud the user deploys to.

Terraform Cost Management#

The most expensive line in your cloud bill was written in a .tf file. A single instance_type choice, a forgotten NAT Gateway, or an over-provisioned RDS instance can cost thousands per month — and none of these show up in terraform plan. Plan shows what changes. It does not show what it costs.

This article covers how to write cost-aware Terraform and catch expensive decisions before they reach production.

Terraform Import and Brownfield Adoption#

Most organizations do not start with Infrastructure as Code. They start with console clicks, CLI commands, and scripts. At some point they decide to adopt Terraform — and now they have hundreds of existing resources that need to be brought under management without disruption.

This is the brownfield problem: writing Terraform code that matches existing infrastructure exactly, importing the state so Terraform knows about the resources, and resolving the inevitable drift between what exists and what the code describes.

Terraform Networking Patterns#

Networking is the first thing you build and the last thing you want to change. CIDR ranges, subnet allocation, and connectivity topology are difficult to modify after resources depend on them. Getting the network right in Terraform saves months of migration work later.

This article covers the networking patterns across AWS, Azure, and GCP — from basic VPC design to multi-region hub-spoke topologies.

CIDR Planning#

Plan CIDR ranges before writing any Terraform. Once a VPC is created with a CIDR block, changing it requires recreating the VPC and everything in it.