Environment_Infrastructure/setup/02-test-terraform-iaac.md

02 - Test Terraform IaC

The purpose of this phase is to create the minimum IaaS resources inside the test Hetzner Cloud Project with Terraform. This document is written so it can be applied on its own.

Scope

Terraform creates the following in the test environment:

• Private network: iklim-test-net
• Subnets:
  • App/Swarm subnet: 10.10.10.0/24
  • DB subnet: 10.10.20.0/24
• Firewall:
  • Public ingress: only 22/tcp, 80/tcp, 443/tcp
  • Private ingress: test rules in 01-private-network-port-matrisi.md
• SSH key
• Placement group: iklim-test-spread
• Floating IP: stable IPv4 for the swarm entry point
• Server:
  • iklim-app-01
  • iklim-db-01
• Ansible inventory output

Terraform does not install DB software. The DB node is prepared only at the machine, network, and firewall level.

Recommended File Structure

terraform/
  hetzner/
    test/
      versions.tf
      providers.tf
      variables.tf
      locals.tf
      network.tf
      firewall.tf
      placement.tf
      servers.tf
      floating_ip.tf
      outputs.tf
      terraform.tfvars.example

terraform.tfvars will not be committed. It must be ignored in .gitignore.

Variables

Minimum variables:

hcloud_token              = "secret"
location                  = "fsn1"
image                     = "rocky-10"
server_type_app           = "cpx42"
server_type_db            = "cpx42"
admin_ssh_public_key_path = "~/.ssh/id_rsa.pub"
admin_allowed_cidrs       = ["X.X.X.X/32"]

The environment constant is in locals.tf; it is not overridden with tfvars.

Start with a single location for location. Disaster recovery across different regions/locations is outside the scope of this stage and must be added to the document later.

The server type decision is based on the current test environment metrics in ../hetzner-sizing-report.md. Because 10 microservices and infrastructure services run together on the test app node, cpx32 was considered risky in terms of RAM. cpx42 is also recommended for the test DB node because of single-node CPU spike risk.

Server Roles

Server	Private IP	Role
iklim-app-01	10.10.10.11	Swarm manager + app worker + Gitea runner
iklim-db-01	10.10.20.11	DB node prepared for manual DB installation

Private IPs must be statically defined inside Terraform. Ansible inventory and firewall rules remain deterministic.

Recommended Resources and Cost

Server	Role	Server Type	CPU	RAM	SSD	Monthly
iklim-app-01	Swarm manager + app worker + Gitea runner	cpx42	8 AMD	16 GB	320 GB	$29.99
iklim-db-01	PostgreSQL/PostGIS + MongoDB node	cpx42	8 AMD	16 GB	320 GB	$29.99
Total	2 servers		16 vCPU	32 GB	640 GB	$59.98

Firewall Rules

Public ingress:

Port	Source	Target
22/tcp	admin_allowed_cidrs	All test nodes
80/tcp	0.0.0.0/0, ::/0	iklim-app-01
443/tcp	0.0.0.0/0, ::/0	iklim-app-01

For public ingress, 8200/tcp, 5432/tcp, 27017/tcp, 5672/tcp, 15672/tcp, 6379/tcp, 2379/tcp, 9000/tcp, 9180/tcp, 9090/tcp, and 3000/tcp will not be opened.

App (swarm) Firewall — Private Ingress

Source from app subnet (iklim-app-01):

Port	Service	Access method
2377/tcp	Docker Swarm control plane	From app subnet
7946/tcp,udp	Docker Swarm node discovery	From app subnet
4789/udp	Docker Swarm VXLAN overlay	From app subnet
8200/tcp	Vault	Docker overlay / private network
6379/tcp	Redis	From app subnet
5672/tcp	RabbitMQ AMQP	From app subnet
61613/tcp	RabbitMQ STOMP	From app subnet
15674/tcp	RabbitMQ Web STOMP	From app subnet
15672/tcp	RabbitMQ Management	From app subnet; external access through SWAG 443 — IP restricted
9000/tcp	APISIX Dashboard	From app subnet; external access through SWAG 443 — IP restricted
9180/tcp	APISIX Admin API	From app subnet, including Docker overlay
9090/tcp	Prometheus	From app subnet; external access through SWAG 443 — IP restricted
3000/tcp	Grafana	From app subnet; external access through SWAG 443 — IP restricted

Source from DB subnet, because iklim-db-01 joins Swarm as a worker:

Port	Service	Source
2377/tcp	Docker Swarm control plane	10.10.20.0/24
7946/tcp,udp	Docker Swarm node discovery	10.10.20.0/24
4789/udp	Docker Swarm VXLAN overlay	10.10.20.0/24

DB Firewall — Private Ingress

Port	Service	Source
22/tcp	SSH	admin_allowed_cidrs
51820/udp	WireGuard VPN	0.0.0.0/0, ::/0 — authentication with cryptographic key
5432/tcp	PostgreSQL	10.10.10.0/24 (app subnet)
27017/tcp	MongoDB	10.10.10.0/24 (app subnet)
2377/tcp	Docker Swarm control plane	10.10.10.0/24 (app subnet)
7946/tcp,udp	Docker Swarm node discovery	10.10.10.0/24 (app subnet)
4789/udp	Docker Swarm VXLAN overlay	10.10.10.0/24 (app subnet)

IP restriction is done in the SWAG nginx configuration, not in the Hetzner firewall. None of these ports are opened publicly from the admin_allowed_cidrs source.

For other private ingress rules, 01-private-network-port-matrisi.md will be used as the source.

Placement Group

The iklim-test-spread placement group will be type = "spread". Because there are two servers in test, this group aims to distribute the iklim-app-01 and iklim-db-01 machines across different physical hosts.

Note: A spread placement group is not a guarantee of a different cabinet or location; it reduces the impact of a single physical host failure.

Terraform Output Expectations

outputs.tf must produce at least the following information:

output "ansible_inventory_yaml" {
  sensitive = false
}

output "test_private_ips" {
  sensitive = false
}

output "test_public_ips" {
  sensitive = false
}

output "test_floating_ip" {
  sensitive = false
}

The inventory output can later be written to ansible/inventory/generated/test.yml. If the inventory file contains no secrets, it can be committed; if it contains secrets or tokens, it will not be committed.

Lifecycle and Resize Policy

server_type Change (Resize)

Changing server_type does not trigger Terraform destroy+create. The hcloud provider supports this natively: it stops the server, calls the Hetzner Resize API, and starts it again. Update the value in terraform.tfvars and run terraform apply.

There is downtime, because the server stops and starts, but disk, installed software, and Docker volumes are preserved. No ignore_changes or manual step is required.

Which Changes Force Server Recreation?

Changed field	Behavior	Note
server_type	In-place resize (provider native)	terraform apply is enough
hcloud_server_network	Only attachment is updated	Because a separate resource is used
hcloud_firewall_attachment	Only attachment is updated	Because a separate resource is used
placement_group_id	Hetzner API does not allow changing it -> destroy+create	Do not change
image	Disk image changes -> destroy+create	Do not change
location	Cannot be moved to another datacenter -> destroy+create	Do not change

Network and Firewall Attachment Separation

The network block and firewall_ids are not embedded inside hcloud_server. Instead, separate resources are defined:

• hcloud_server_network — private IP assignment
• hcloud_firewall_attachment — firewall relationship

In embedded definitions, some provider versions interpret changes in these fields as server recreation. When separate resources are used, only the attachment is updated and the server is left untouched.

prevent_destroy Protection

Each server gets lifecycle { prevent_destroy = true }. While this block exists, Terraform cannot delete the server under any condition and fails during the plan phase. To intentionally delete a server, temporarily remove the lifecycle block first.

Acceptance Criteria

• terraform plan works only with the test Hetzner Project token.
• 2 servers are created after terraform apply.
• The two servers can reach each other through the private network.
• Only 22, 80, and 443 are open at firewall level from the public internet.
• Vault 8200 remains closed from the public internet.
• Terraform state is not committed to the repo.