Implementation Guides

Chapter 7: Implementation Guides

This chapter translates the architecture patterns from Chapter 6 and the tooling decisions from Chapter 5 into working implementation. It is the most technically detailed section of the handbook — written for the engineers and architects who must build what the earlier chapters describe.

The chapter is structured as five implementation guides, each addressing a distinct layer of the automation capability stack. They are designed to be read in order for a new implementation, or individually when working on a specific area.

Guide	What It Covers
Config as Code	Building the source of truth, Jinja2 templates, and Ansible playbooks
CI/CD Pipelines	End-to-end pipeline design across nine stages
Testing Strategies	The layered testing taxonomy, from lint to live device verification
Deployment Patterns	Replace vs merge, canary, rollback, and blast radius containment
One-Touch Deployment	Intent-driven generation, guardrails, and self-service provisioning

---

How These Guides Relate to the Architecture

Each guide implements a specific part of the architecture introduced in Chapter 6.

graph LR
    subgraph "Chapter 6 — Architecture"
        CaC["Config as Code Pattern"]
        IBN["Intent-Based Architecture"]
        REPO["Repository Structure"]
    end

    subgraph "Chapter 7 — Implementation"
        G1["Config as Code Guide"]
        G2["CI/CD Pipeline Guide"]
        G3["Testing Guide"]
        G4["Deployment Guide"]
        G5["One-Touch Guide"]
    end

    CaC --> G1
    CaC --> G2
    IBN --> G3
    IBN --> G5
    REPO --> G1
    G1 --> G2
    G2 --> G3
    G2 --> G4
    G4 --> G5

View full size

The guides build on each other. Config as code provides the source of truth and templates that the CI/CD pipeline consumes. The pipeline stages include the testing layers described in the testing guide. Deployment patterns are pipeline stages. One-touch deployment is the highest-order capability, built on all the layers below it.

---

The ACME Investments Implementation

Throughout this chapter, ACME Investments’ implementation is used as the running example. By the end of the five guides, you will have a complete picture of a working network-as-code implementation:

• requirements.yml → design_intents.yml → nodes.yml — the three-layer intent model
• templates/arista_eos/ and templates/cisco_ios/ — multi-vendor template library
• verify_intents.py + batfish_validate.py — the two-layer verification suite
• .gitlab-ci.yml — a nine-stage pipeline from lint to deployment
• generate_branch.py — one-touch branch provisioning with guardrails

This is not a proof of concept. It is a production-representative implementation that can serve as the foundation for a real programme.

---

A Note on Depth and Audience

These guides include code examples and configuration fragments — YAML, Python, and pipeline definitions. The intent is architectural illustration, not step-by-step tutorial. The examples show what the implementation looks like and why it is structured the way it is. For the full working codebase, see the examples/acme-investments-demo/ directory.

Engineers new to specific tools (Ansible, Batfish, Jinja2) should use the tool’s own documentation alongside these guides. The handbook does not replicate what is already well-documented elsewhere.

---

Begin with: Config as Code