/tf-cost

@rules/infrastructure/STRUCTURE.md @rules/infrastructure/MODULES.md

Terraform Cost Estimation: {{ args[0] | default: "infrastructure review" }}

Review infrastructure costs and provide sizing recommendations.

Step 1: Identify Resources

# List all resource types in the current config
grep -r "^resource " *.tf modules/ 2>/dev/null | awk '{print $2}' | sort | uniq -c | sort -rn

Step 2: Common Cost Patterns

RDS (PostgreSQL/Aurora)

Environment	Instance	Storage	Multi-AZ	Est. Monthly
dev	db.t3.micro	20 GB gp3	No	~$15
staging	db.t3.small	50 GB gp3	No	~$30
prod	db.r6g.large	100 GB gp3	Yes	~$400

Optimization:

• Use db.t3.* for dev/staging — burstable is fine for low traffic
• Enable storage autoscaling with a max limit
• Aurora Serverless v2 for variable workloads (0.5-128 ACU)

ElastiCache (Redis)

Environment	Instance	Replicas	Est. Monthly
dev	cache.t3.micro	0	~$12
staging	cache.t3.small	0	~$25
prod	cache.r6g.large	1	~$300

NAT Gateway

Cost trap: NAT gateways charge per hour AND per GB processed.

Setup	Est. Monthly
1 NAT per AZ (3 AZs)	~$100 + data
1 shared NAT	~$33 + data
NAT instance (t3.micro)	~$8

Optimization:

• Dev: use a single NAT or NAT instance
• Prod: one NAT per AZ for high availability
• Route S3/DynamoDB through VPC endpoints (free) instead of NAT

ECS Fargate

Size	vCPU	Memory	Est. Monthly (24/7)
Micro	0.25	0.5 GB	~$9
Small	0.5	1 GB	~$18
Medium	1	2 GB	~$36
Large	2	4 GB	~$73

Optimization:

• Use Fargate Spot for dev/staging (up to 70% savings)
• Right-size: check CloudWatch CPU/memory utilization
• Scale to zero in dev after hours

EKS Node Groups

Environment	Instance	Nodes	Est. Monthly
dev	t3.medium	2	~$60 + $75 (control plane)
prod	m6i.xlarge	3-6	~$300-600 + $75

Note: EKS control plane is $75/month regardless of size.

Step 3: Dev vs Prod Sizing

Resource	Dev	Prod	Savings
RDS	t3.micro, no Multi-AZ	r6g.large, Multi-AZ	95%
Redis	t3.micro, no replica	r6g.large, 1 replica	95%
NAT	1 shared or NAT instance	1 per AZ	70%
Fargate	Spot, min replicas	On-demand, auto-scale	60%
EKS	Spot nodes, smaller	On-demand, right-sized	50%

Step 4: Cost Optimization Tips

1. VPC Endpoints — S3 and DynamoDB gateway endpoints are free. Saves NAT data costs.
2. Reserved Instances — 1-year no-upfront saves ~30% on RDS/ElastiCache in prod.
3. Scheduled scaling — Scale down dev/staging outside business hours.
4. S3 lifecycle rules — Move old objects to Glacier or Intelligent-Tiering.
5. Right-size — Review CloudWatch metrics monthly. Downsize over-provisioned resources.
6. Clean up — Delete unused EBS snapshots, old AMIs, unattached EIPs.

Step 5: Estimate for Current Config

Review the Terraform config and produce:

## Cost Estimate: {env}

| Resource | Type | Size | Est. Monthly |
|----------|------|------|-------------|
| RDS | PostgreSQL | db.t3.small | $30 |
| ... | ... | ... | ... |

**Total estimated:** $X/month
**Potential savings:** $Y/month with [recommendations]

Rules

• These are estimates — use AWS Pricing Calculator for exact numbers
• Always compare dev vs prod sizing — dev should be minimal
• Flag NAT gateway costs explicitly — they surprise teams
• Recommend VPC endpoints before adding NAT gateways for AWS service traffic
• Cost reviews should happen before deploying new infrastructure, not after