641 words
3 min read

GitOps on AWS — Real-World DevOps Pipeline with Argo CD, Terraform & EKS

By · Docker Captain · CNCF Ambassador · IBM Champion
Wooden desk with dual ultrawide monitors showing a minimalist "DO MORE" wallpaper, a tidy mechanical keyboard, open notepad, large potted monstera, and a shelf with retro license plates

This is how I built a production-grade GitOps pipeline on AWS. Fully containerized, drawn straight from what I run in production.

No buzzwords. No clickbait. The architecture works, and the mindset behind it holds up as you grow.

Why I Chose GitOps#

GitOps is about control. Git becomes the one place that tells the truth about your system.

  • Every change goes through a pull request.
  • Every rollback is just a git revert.

No guesswork. Nobody asking who deployed what on Friday night. By 2025 this stopped being a trend. If your team is serious about infrastructure, it’s the floor you start from, not the ceiling you reach for.

So why does my setup look the way it does? Start with the stack.

My Stack at a Glance#

One terraform apply. That gets you a reproducible platform, codified end to end, with nothing done by hand.

Why Containers?#

A container is the smallest unit of reliability I know of. It runs the same way in dev, in staging, and in production. Isolated, predictable, pinned to a version. CI builds the image and tags it, say release-2025.04.16-prod, then pushes it to Amazon ECR. What runs in production is that exact artifact, byte for byte.

You know the old joke about how it works on my machine? Containers retire it. You ship systems instead of chaos.

The GitOps Mindset Shift#

GitOps asks you to rethink one thing, and it trips up plenty of experienced teams. Here’s the mistake I keep running into.

They assume CI should handle deployments. Under GitOps it doesn’t.

  • CI’s job is simply to push changes to Git.
  • Argo CD handles the deploy. On its own. On schedule. No manual triggers.

That split is the whole point. Git holds the truth. CI is just the courier that gets your change there.

How It All Connects#

  • CI runs on GitHub Actions.
  • It builds the Docker image, pushes it to Amazon ECR, updates Helm values, and commits to Git.
  • Argo CD detects changes and applies them to the cluster.
  • Terraform provisions the entire platform, Argo CD included.
  • Vault integrates securely, providing secrets at runtime.

So there are no plain-text tokens sitting around. No unencrypted environment variables either.

This stack does the job, and it holds up when something goes wrong.

Hard-Earned Lessons#

Getting here took a while, and I paid for a few of these the hard way. Here’s what I’d tell you so you skip my mistakes.

1. Bootstrapping#

Argo CD doesn’t magically install itself.

You need a plan for that first install. I use Terraform and the Helm Provider to automate it.

2. Namespaces#

Never run Argo CD alongside your applications.

Keep it isolated. Your future self will thank you for it.

3. Secrets#

If your secrets live in YAML files, that isn’t GitOps. That’s hopeOps.

Use Vault or AWS Secrets Manager. Credentials never go in the open.

Monitoring: The Non-Negotiable#

If your monitoring is users calling you at 3 AM, that isn’t monitoring. It’s a nightmare with a phone number.

For metrics I run Prometheus and Grafana. Loki handles logs, Alertmanager handles alerts. Argo CD exposes its own metrics too, so the moment the cluster drifts from Git, I see it.

Monitoring isn’t a bolt-on you add later. Skip it and you’re flying blind.

What Success Looks Like#

  • Deployment time: minutes, not hours.
  • Rollbacks: one click.
  • New environments: one command.
  • New developers: clone and go.
  • Everything documented, repeatable, and under control.

None of that is hype. That’s the day-to-day.

Final Thoughts#

This is more than a list of tools. It’s a way of working. Containers, infrastructure as code, every change routed through Git: that’s what a mature system looks like in 2025. GitOps was never really about YAML. It’s about building something you can actually trust.

Thank you for reading! Don’t forget to check out the video version for additional insights and visuals.


Vladimir Mikhalev

Docker Captain  ·  CNCF Ambassador  ·  IBM Champion

The Verdict — production-tested analysis on YouTube.

The Verdict

Inconvenient truths about shipping in the AI era

Container security, platform engineering, and the agentic shift — tested in production, argued without the hype. The verdict reaches your inbox the moment there's one worth sending.

Related Posts

Same category
  1. 1
    EKS Validation on a Real AWS Account: Twenty Cents a Run, Seven Failures No Mock Saw
    DevOps & Cloud · Proving an EKS platform on a real AWS account cost twenty cents a run. Seven failed runs before it found what mocked Terraform tests and kind never could.
  2. 2
    A Partial Read Is Not a Smaller Fleet: A GitHub API Rate Limit and the Numbers That Lied for an Hour
    DevOps & Cloud · A partial read published as the whole. For an hour the profile said 24 restore scripts across 15 repositories, and the fleet had 73 across 48. Three guards now.
  3. 3
    Restore Drills on a Clean Machine — Measured RTO for 47 Stacks, 19 Seconds to 7 Minutes
    DevOps & Cloud · Restore drills that measure RTO on a machine that never ran the stack, 47 templates weekly. Every same-host restore passed. The first clean one answered 500.
  4. 4
    Mutation Testing for Infrastructure Code — 153 Planted Bugs, Four Tests That Could Not Fail
    DevOps & Cloud · Mutation testing for infrastructure code, done by hand across fifteen repositories. 153 planted bugs, and four green tests that could never have failed.

Random Posts

Random
  1. 1
    watsonx.governance Enforcement Tracking — Sampled Compliance and the Unevidenced Run
    AI & MLOps · EU AI Act Article 12 wants a per-event record. Enforcement Tracking evidences thresholds on a schedule. What that gap costs, and the join key that closes it.
  2. 2
    Your Knowledge Is a Depreciating Asset. Judgment Compounds.
    Opinion & Culture · AI made reproducible knowledge free, so technical expertise is now a depreciating asset. Judgment is the one that compounds. Here is how to move your weight.
  3. 3
    Install Minecraft Server Using Docker Compose
    Self-Hosting · Learn how to install a Minecraft Server using Docker Compose. Set up your own secure multiplayer server on Ubuntu with ease using this step-by-step guide.
  4. 4
    Install Gitea Using Docker Compose
    Self-Hosting · Learn how to install Gitea using Docker Compose and Traefik with free SSL from Let's Encrypt. Self-host your Git service with secure Git over SSH.
GitOps on AWS — Real-World DevOps Pipeline with Argo CD, Terraform & EKS
https://heyvaldemar.com/gitops-on-aws-real-world-devops-pipeline-with-argo-cd-terraform-and-eks/
Author
Vladimir Mikhalev
Published
2025-04-22
License
CC BY-NC-SA 4.0