Capstone: Full Production Deployment

Throughout this Sub-module, you've built cloud deployment knowledge step by step: cluster provisioning with doctl and hetzner-k3s, load balancer configuration, stack deployment with Dapr and Traefik, secrets management, and production verification. Now it's time to apply everything to a complete production scenario.

In previous chapters, you deployed components individually. You tested pieces in isolation. But production deployment isn't about individual commands working in isolation. Production deployment is about the complete sequence: from empty cloud account to working HTTPS endpoint serving real traffic.

This capstone brings it all together. You'll write a deployment specification FIRST, provision real infrastructure, deploy your complete stack, verify with HTTPS, and execute a clean teardown. The specification-first approach is what separates professional cloud engineering from Vibe Coding.

The result: your multi-cloud-deployer skill becomes production-ready. It's not a theoretical exercise. It's a verified, tested Digital FTE component you can use for real deployments.

---

Phase 1: Write the Deployment Specification

Before touching any CLI tools, write your specification. This forces you to think about environment choices, resource requirements, success criteria, and cost constraints before any infrastructure exists.

Create this file in your project directory:

File: deployment-spec.md

Why specification first?

Without a spec, you'd start running doctl kubernetes cluster create and figure things out as you go. That's Vibe Coding. You might choose the wrong node size. You might forget TLS configuration until you're debugging certificate errors at 2 AM. You might deploy to the most expensive region.

The spec makes constraints explicit BEFORE you provision. It's your contract with yourself and your budget.

---

Phase 2: Choose Your Deployment Path

This capstone supports three deployment paths. Each path produces identical outcomes with different providers and cost profiles.

Path A: DigitalOcean DOKS (Production Path)

Best for: Teams, real traffic, managed SLA

Monthly cost: ~$48+ (3-node cluster minimum)

Provisioning command:

Output:

Connect and verify:

Output:

---

Path B: Hetzner + K3s (Budget Path)

Best for: Personal practice, budget-conscious learners

Monthly cost: ~$15 (3x CX22 servers)

Provisioning command:

Output:

Connect and verify:

Output:

---

Path C: Azure AKS (Enterprise Path)

Best for: Enterprise environments, Azure ecosystem integration

Monthly cost: ~$75+ (3-node cluster)

Provisioning command:

Connect and verify:

---

Phase 3: Deploy the Complete Stack

With your cluster provisioned, deploy the complete production stack. This sequence is universal across all providers.

Step 1: Install Dapr

Output:

Verify:

Output:

Step 2: Install Traefik Ingress

Output:

Get the Load Balancer IP:

Output:

Record this IP for DNS configuration.

Step 3: Install cert-manager

Output:

Step 4: Configure Let's Encrypt ClusterIssuer

File: cluster-issuer.yaml

Apply:

Output:

Step 5: Deploy Task API

File: task-api-deployment.yaml

Apply:

Output:

Step 6: Configure Ingress with TLS

File: task-api-ingress.yaml

Apply:

Output:

---

Phase 4: The Convergence Loop

Real deployments rarely work perfectly on the first try. This phase demonstrates the iterative refinement process that produces working production systems.

Iteration 1: Initial Deployment Attempt

After applying all resources, check status:

Common issues and their indicators:

Iteration 2: Troubleshoot and Fix

For image pull issues:

If you see "unauthorized," create an image pull secret:

For DNS issues:

Point your domain's A record to the Load Balancer IP. Verify propagation:

Output:

For certificate issues:

Check cert-manager logs:

Iteration 3: Verify Success

When all issues are resolved:

Output:

Output:

Output:

---

Phase 5: Production Verification Checklist

Go back to your specification and verify each success criterion:

---

Phase 6: Complete Teardown with Cost Verification

Production capstones aren't complete without teardown. You must prove zero ongoing costs.

For DigitalOcean DOKS

Output:

For Hetzner + K3s

Output:

For Azure AKS

Cost Verification

Check your provider dashboard:

1. Log in to your cloud provider's billing console
2. Verify no active resources
3. Confirm next billing cycle shows $0 for this project

Update your specification checklist:

All criteria met. Specification satisfied.

---

Final Skill Evaluation Rubric

Your multi-cloud-deployer skill has been tested and refined throughout this Sub-module. Evaluate its production readiness:

Total: /100

Scoring guide:

• 90-100: Your skill is production-ready and sellable as a Digital FTE component
• 70-89: Minor gaps to address, but functional for personal use
• 50-69: Significant improvements needed before production use
• Below 50: Return to earlier chapters and rebuild skill foundations

---

Reflect on Your Skill

Your multi-cloud-deployer skill has evolved through this Sub-module. It started as a skeleton created from official documentation. Now it's been tested against real cloud infrastructure.

Final Test: Ask your skill:

Evaluate the output:

• Does it include all provisioning commands (doctl kubernetes cluster create)?
• Does it include the deployment sequence (Dapr -> Traefik -> cert-manager -> app)?
• Does it include verification steps (kubectl get pods, curl https://...)?
• Does it include teardown commands (doctl kubernetes cluster delete)?

Your skill is production-ready when:

• It generates accurate CLI commands for at least 3 providers (DOKS, Hetzner, AKS)
• It includes the universal "provision -> connect -> deploy" pattern
• It can answer questions about cost and tradeoffs between providers
• It produces deployment sequences that work on first try (or identifies likely issues)

This skill is now part of your Digital FTE portfolio.

You don't just "know cloud deployment." You OWN a verified, production-tested skill that can deploy AI agent services to any major cloud provider. This is the outcome of the Skill-First Learning Pattern: not knowledge, but assets.

---

Try With AI

You've completed the capstone by following the specification-first approach. Now extend your deployment skills through AI collaboration.

Prompt 1: Specification Review

What you're learning: AI can review specifications and identify blind spots. It might suggest failover strategies, backup procedures, or monitoring configurations you hadn't considered. You evaluate each suggestion against your actual production requirements.

Prompt 2: Multi-Cloud Comparison

What you're learning: The "provision -> connect -> deploy" pattern is universal, but provisioning commands differ. AI helps you understand which skills transfer directly and which require adaptation.

Prompt 3: Production Hardening

What you're learning: A working deployment isn't a hardened deployment. AI helps you identify the gap between "it runs" and "it's production-ready." You evaluate each recommendation against your application's actual risk profile.

Safety note: When sharing deployment specifications with AI, redact actual domain names, IP addresses, and cloud credentials. Replace real values with placeholders like yourdomain.com or YOUR_API_TOKEN.