OpenTofu and Ansible: infrastructure and configuration as code

This lesson starts from zero. An unfamiliar word is not treated as your deficiency: we first build an image, then give it a precise meaning, and only then use it.

Where we are

Flight review. The previous lesson used the support domain → DNS → public 443 → Caddy → private app:8080; certificate proves name. Without looking, name its four links, then compare with the written signal.

Lesson outcome

Validate a resource description, produce inventory, and apply idempotent host configuration. Do not move on until you can explain what every command verifies and which failure it can reveal.

Begin with a familiar image

A builder works from a drawing while a manager keeps a ledger of what already exists. OpenTofu describes desired cloud resources and links them to real objects in state. Ansible resembles a room setup guide that installs packages and brings files to a desired form.

An analogy starts reasoning but does not replace the technology. We make every word precise below.

The lesson’s support signal

desired code → plan/diff → approve → apply → state; repeat → no surprise

Read the chain left to right. It answers not “which command should I memorise?” but “why does each step follow the previous one?”

New words in plain language

• Infrastructure as Code — Infrastructure described in reviewable, versioned files.
• State — A sensitive record mapping code to real resources.
• Plan — A preview of proposed changes before application.
• Idempotence — Repeating an operation after reaching the desired state makes no needless changes.

Take it apart without rushing

OpenTofu manages infrastructure resource lifecycles through state; Ansible brings the OS to the desired configuration over SSH. Code makes changes reviewable, but state and secrets need protection. An idempotent playbook should report no changes on a second run.

Do not memorise this wording. Find its cause-and-effect link and attach it to the support signal above.

Predict the result first

Before running anything, write what should change, what should remain unchanged, and which output would support your prediction. A wrong prediction is useful when you can explain the difference after observing it.

Experiment in small steps

Run the commands from the course/cloud-devops-lab root. The commands target a learning environment. Read a command in full and check the current directory before it changes a server or data. Run one meaningful group at a time. Stop after each group and compare evidence with your prediction.

tofu -chdir=infra/opentofu fmt -check 2>/dev/null || true
tofu -chdir=infra/opentofu init -backend=false 2>/dev/null || true
tofu -chdir=infra/opentofu validate 2>/dev/null || true
ansible-playbook --syntax-check -i infra/ansible/inventory.ini.example infra/ansible/site.yml 2>/dev/null || true

Project reference files.

What this experiment actually does

fmt checks form, init -backend=false prepares local validation without remote state, and validate checks structure. Ansible syntax-check reads the playbook without changing a server. No lesson command invokes apply.

What you should observe

• Validation creates no cloud resource.
• The command exits with a predictable status.
• The result can be repeated in a clean environment.

Rebuild the whole from the support map

Follow the four blocks with a finger or pencil and tell the entire lesson as one connected explanation.

Recall without a prompt

1. Hide the map and draw its four blocks from memory.
2. Explain every transition using the word “because”.
3. Define one new term without repeating the text verbatim.
4. Name one signal that proves the result.

Find and correct the mistake

Mistake: “Infrastructure code is safe because it is readable.” Correction: wrong code repeats mistakes faster; use plans, review, limited rights, and protected state.

Exercise

Now rebuild the same meaning yourself. You may use the support map; the goal is a repeatable result rather than a memory exam.

Required. Write a POSIX sh preflight that requires tofu and ansible-playbook, runs fmt/validate and syntax-check, and never calls apply. Submit only POSIX sh commands or a script, with no real keys or tokens.

Completion criteria

• You understand every line you run.
• A repeat gives the expected result or a safely explained difference.
• Output contains no real keys, tokens, or passwords.
• You can point to evidence that proves success.

If a check fails, record expected and actual results, revisit the one broken support link, and retry. An error does not remove your right to continue learning.

Optional. Write down one possible failure and the signal that would reveal it.

Open perspective

The next lesson adds a new support: Kubernetes: Deployment, Service, probes, and rollout.

Course contents