↓ Skip to main content

Breaking into DevOps: What to Learn, What to Build, and How to Show It

888 words·5 mins· loading · loading ·
Table of Contents
Mysteries of DevOps - This article is part of a series.
Part 4: This Article

Most advice about getting into DevOps is a list of tools: learn Git, learn Docker, learn Kubernetes, learn Terraform, learn a cloud. The list isn’t wrong. The problem is that finishing it doesn’t make you employable, because everyone applying has the same list, and the same certificates and tutorial projects to show for it.

What sets people apart is evidence that they’ve run something real, watched it break, and fixed it properly. This post covers the order I’d learn things in, why one long-running project beats a pile of tutorials, and how to make that work visible.


What you already have
#

If you’re coming from another field, some of what DevOps work needs is already there. It just isn’t labelled DevOps yet.

If you’ve done… You already practise…
IT support or helpdesk Troubleshooting under pressure, users, tickets, Windows and networks
System or network administration Most of the operations side: servers, DNS, firewalls, backups
Software development or testing The development side: code, version control, tests
QA, analysis or project work Process thinking, requirements, finding where handoffs fail
Anything with on-call or shifts Incident response, and knowing what a 3 a.m. alert feels like

Name these on your CV in DevOps terms. “Kept a 40-person office’s network running” is operations experience.


An order that works
#

The map of DevOps areas in the previous post lists ten things. This is the order I’d learn them in, and roughly why.

  1. Linux and the command line. Everything else runs on it. Get comfortable with files, permissions, processes, services and logs.
  2. Networking basics. IP, DNS, ports, NAT and TLS. Most “mysterious” failures are one of these.
  3. Git. Daily, for everything, including notes and config. Commits, branches and pull requests.
  4. Bash, then Python. Enough to automate a task you do by hand. (Is coding essential? The next post answers that.)
  5. One cloud. Pick one of AWS, Azure or GCP and learn compute, networking, storage and identity there. The concepts transfer.
  6. Docker. Package an app you understand, run it, break it, rebuild it.
  7. CI/CD. One pipeline that tests and deploys on every push. GitHub Actions is the easiest place to start.
  8. Infrastructure as code. Terraform to create the servers, Ansible to configure them.
  9. Monitoring. Metrics, dashboards, alerts and logs, and an alert that reaches your phone.
  10. Kubernetes. Last, on purpose. It’s far easier to learn once you’ve done by hand what it automates.

Tools on this list will change. In 2023 I’d have put Chef and Puppet next to Ansible. Today I wouldn’t start with them. Don’t worry about picking the “right” tool. Worry about understanding the problem it solves, because that part carries over.


Build one thing and run it for months
#

Tutorial projects all look the same: deploy an app to Kubernetes, screenshot the dashboard, done. They show you can follow instructions. They don’t show what happens next, which is most of the job.

Build one thing you’ll actually use, and keep it running for months. A personal website, a small app, a home server. Then add to it as you learn: put it in Git, write a deploy script, add a pipeline, define the server in Terraform, add monitoring and backups.

The value comes when it breaks, and it will. My own homelab, Project Aether, taught me more in its failures than in its setup:

  • backups that reported success for months while never being checked
  • a “quick fix” to monitoring that exposed my admin tools to the wrong machine
  • a cluster tool that quietly drifted from what Git said it should be

None of those would have shown up in a weekend project. All of them are the kind of problem DevOps teams deal with every week, and being able to talk through one in an interview is worth more than any certificate.


Make the work visible
#

A hiring manager can’t see what’s on your laptop. Give them something to check.

  • A public repository. Real code and config, a README that explains what it does and why, and a commit history that shows how it grew. Keep secrets out of it (encrypted, never plaintext).
  • Written decisions. Short notes on why you chose one tool over another, and what you gave up. This shows judgement, which is hard to show any other way.
  • Incident write-ups. When something breaks, write down what happened, the timeline, the root cause and what you changed. No blame, including on yourself. This is one of the strongest signals you can give, because it’s exactly what senior engineers do.
  • A few blog posts. Explaining something forces you to understand it. Write about the problem you got stuck on, not the tutorial you followed.

The parts that aren’t technical
#

Two more things matter more than most lists admit.

Talk to people who do the job. Online communities and local meetups are where you hear what the work is actually like, and where job leads come from. Ask specific questions about their systems. People like talking about their systems.

Keep going after it gets boring. The first month of a project is exciting. The value is in month six, when it’s routine and something breaks anyway. DevOps is mostly that: keeping systems boring, and fixing them well when they aren’t.

The next post answers the question most people ask before they start: do you need to code?

Mysteries of DevOps - This article is part of a series.
Part 4: This Article