Implement State File Separation for Infrastructure Components

[runatyr1]

Why Not Workspaces

• Less explicit dependency management
• Higher risk of accidental state modifications
• Harder to manage different configurations per component
• More difficult to handle concurrent modifications

Proposed Structure

terraform/
├── environments/
│   └── staging/
│       ├── core/           # VPC, IAM, EKS (required first)
│       ├── apps/          # ArgoCD, other k8s deployments
│       └── data/          # RDS, S3, ElastiCache (future)

Changes Required

New Files

1. staging/core/main.tf

• Move VPC, EKS from current staging/main.tf
• Add data outputs for apps/data dependencies

2. staging/core/outputs.tf

output "eks_cluster_endpoint" {}
output "vpc_id" {}
output "private_subnets" {}

3. staging/apps/main.tf

• Move ArgoCD configuration
• Add data sources for core dependencies:

data "terraform_remote_state" "core" {
  backend = "s3"
  config = {
    bucket = "your-bucket"
    key    = "staging/core/terraform.tfstate"
  }
}

4. staging/data/main.tf (future)

• Similar core dependency pattern
• Separate state for databases/storage

Modified Files

1. Move provider configs to each component
2. Update backend configurations for state separation
3. Update module references to use remote state data

Dependencies

graph TD
    Core --> Apps
    Core --> Data

Loading

• Core must be deployed first
• Apps and Data can be deployed independently
• Each component gets its own terraform.tfstate

Migration Steps

1. Backup current state
2. Create new S3 backend paths
3. Move resources to new structure
4. Import existing resources to new states
5. Update CI/CD for separate apply stages

Future Considerations

• IAM roles/policies in core
• Networking components isolated in core
• Monitoring stack could be separate component
• Consider secrets management separation