A VPC with public and private subnets across up to three availability zones, an internet gateway, route tables and an S3 gateway endpoint. A NAT gateway is optional and off by default.
Network only: no compute, no database. It is the plain building block for workloads that you deploy afterwards. OpenTofu demo template. Standalone, no dependency on other repos. Compatible with Terraform.
flowchart TB
internet([Internet]) <--> igw[Internet gateway]
subgraph vpc[VPC 10.0.0.0/16]
subgraph pub[Public subnets, one per AZ]
nat[NAT gateway, optional]
end
subgraph priv[Private subnets, one per AZ]
workload[Your workloads]
end
end
igw <--> pub
priv -->|0.0.0.0/0, only if NAT is on| nat
priv -->|gateway endpoint| s3[(S3)]
vpc -->|flow logs| cw[(CloudWatch Logs)]
- VPC with
az_countpublic andaz_countprivate subnets (default 2 each,/24), public subnets first - Internet gateway and one public route table with a default route to it
- One route table per private subnet. Without the NAT gateway the private subnets have no route to the internet
- NAT gateway (optional): one shared gateway, or one per AZ with
nat_gateway_per_az - S3 gateway endpoint on the private route tables (no hourly cost)
- VPC flow logs (all traffic) to CloudWatch Logs
- Default security group without any rules, default network ACL unchanged
- Subnets do not assign public IPs automatically
- The default security group has no rules, so it allows no traffic
- Private subnets have no route to the internet unless the NAT gateway is on
- Flow logs on by default, retention configurable
vpc_cidrmust be a private range (RFC 1918) between/16and/24, and the module checks that the subnets fit
Accepted checkov findings (documented inline in the code):
| Check | Reason |
|---|---|
| CKV_AWS_338, CKV_AWS_158 | Log retention is configurable, AWS-managed encryption for logs |
| CKV_AWS_394 | Only the first az_count zones are used |
The Trivy scan (trivy config) accepts findings with #trivy:ignore comments above the resource. The reason is written in the same line.
Configure AWS credentials, for example with aws configure sso or environment variables.
cp terraform.tfvars.example terraform.tfvars
tofu init
tofu plan
tofu applyPlace workloads in private_subnet_ids. Use public_subnet_ids for resources that must be reachable from the internet, such as load balancers. Public subnets only have a route to the internet gateway. A public IP address still has to be assigned explicitly. Turn on enable_nat_gateway when private workloads need outbound internet access.
| Name | Description | Default |
|---|---|---|
name_prefix |
Prefix for all resource names | "demo" |
region |
AWS region | "eu-central-1" |
advanced_settings |
Object with the settings below, all optional | {} |
advanced_settings attributes (shown as "Advanced Settings" in the StackGuardian form):
| Name | Description | Default |
|---|---|---|
vpc_cidr |
CIDR block of the VPC. Private range (RFC 1918), prefix /16 to /24 |
"10.0.0.0/16" |
az_count |
Number of availability zones, 1 to 3. Each gets a public and a private subnet | 2 |
subnet_newbits |
Bits added to the VPC prefix for each subnet, 1 to 12. 8 gives /24 subnets in a /16. The subnets must fit: 2 x az_count blocks, and the subnet prefix may not be longer than /28 |
8 |
enable_nat_gateway |
NAT gateway for the private subnets | false |
nat_gateway_per_az |
One NAT gateway per AZ instead of one shared. Needs enable_nat_gateway |
false |
enable_flow_logs |
VPC flow logs to CloudWatch Logs | true |
flow_log_retention_days |
Retention of the flow logs in CloudWatch | 90 |
tags |
Additional tags for all resources | {} |
| Name | Description |
|---|---|
vpc_id |
VPC ID |
vpc_cidr |
CIDR block of the VPC |
public_subnet_ids |
IDs of the public subnets |
private_subnet_ids |
IDs of the private subnets |
public_subnet_cidrs |
CIDR blocks of the public subnets |
private_subnet_cidrs |
CIDR blocks of the private subnets |
availability_zones |
Availability zones of the subnets, in subnet order |
internet_gateway_id |
ID of the internet gateway |
nat_gateway_ids |
IDs of the NAT gateways, empty when off |
private_route_table_ids |
IDs of the private route tables |
The VPC, subnets, route tables, internet gateway and the S3 gateway endpoint have no hourly cost. Each NAT gateway (and its Elastic IP) is billed per hour while it exists, plus data processing, even when idle, so it is off by default. With nat_gateway_per_az you pay for one per AZ. Flow logs are billed by CloudWatch Logs for ingestion and storage.
Destroy after the demo:
tofu destroyThe GitLab pipeline (.gitlab-ci.yml) lints every push. Merge requests into development or main also run
validation, the checkov and Trivy scans and the unit tests. A tag runs all of these plus the integration tests, which deploy
real resources and destroy them again, and then creates a release. Integration tests run only when the CI/CD variable AWS_ROLE_ARN is set: an IAM role trusted for GitLab OIDC with the audience sts.amazonaws.com. AWS_REGION is optional.
In GitLab, the unit-tests job also publishes its results as a JUnit test report, which you find in the Tests tab of the pipeline and of the merge request. The script ci/tofu-json-to-junit.py converts the output of tofu test -json to JUnit XML.
The GitHub Actions workflow (.github/workflows/ci.yml) runs the same jobs: every push, pull requests into development or main, and tags. Set AWS_ROLE_ARN (and optionally AWS_REGION) as repository variables; the IAM role then needs a trust for GitHub OIDC (token.actions.githubusercontent.com).
tofu test -test-directory=tests/unitTest cases: tests/README.md
pre-commit install
pre-commit run --all-files