Ship It

Most teams don't have a deployment problem. They have a drift problem.

The cluster runs fine until nobody can answer what, exactly, is running on it. The repo says one thing, the cluster says another, and the gap grows a little wider every time someone fixes something by hand at 2am.

Ship It: Argo is a hands-on handbook for platform engineers who want to close that gap permanently. It teaches the full Argo ecosystem (Argo CD, Argo Rollouts, Argo Workflows, and Argo Events) not as four separate tools, but as one coherent system: a self-driving delivery platform where Git is the spine and everything else is automation built around it.

What you'll build

Every chapter adds to a single running service called Snapshot, a real image-processing API that genuinely needs all four tools. By the end, a developer's git push triggers a build, a metric-judged canary release, and a fully automated rollback if things go wrong. No human in the loop on the happy path. No duct tape holding it together.

What's covered

- The GitOps idea: why pull beats push, why drift is a disease, and what it actually costs to fix it
- Argo CD: declarative deployment, self-healing, Kustomize overlays, app-of-apps, ApplicationSets, and a clear-eyed treatment of the promotion problem and secrets management
- Argo Rollouts: canary releases with real traffic splitting, automated analysis against Prometheus metrics, and blue-green for when gradual exposure is the wrong shape
- Argo Workflows: DAG-based pipelines, parallel steps, artifact passing, and the write-back pattern that commits image tags to Git without breaking your security model
- Argo Events: webhook sources, the EventBus, Sensors with filters and parameter extraction, and how a GitHub push becomes a production deploy with no human relay
- Production: RBAC, observability, scaling, and the anti-patterns that bite teams who skip Part 6

What makes this different

The book is honest. It tells you the rough edges: the promotion problem the docs skip, the self-heal gotcha that turned one team's on-call shift into a forty-minute fight with their own tooling, the secrets trap that bites everyone eventually. It doesn't pretend GitOps is easy or that the tools are magic. It shows you how they actually work, where they fit together, and where to be careful. You can stop at the end of any part and have a working system. If all you ever want is GitOps deployment, Part 2 is a complete book.

Who this is for

Platform engineers, SREs, and senior developers who run Kubernetes in production and want their delivery pipeline to be a system they trust at 3am, not a series of manual steps held together by tribal knowledge.

Requirements: Kubernetes basics, a cluster you can break (local is fine), and a willingness to install a CLI or two as you go.