The minimum setup that tells you what broke before your users do. The best deploys are the ones nobody notices. Everything in this post exists to make releases small, reversible, and dull enough that we can ship on a Friday afternoon.

Keep production boring
Our pipeline runs the same steps locally and in CI: install, lint, test, build, and publish an image tagged with the commit hash. If a step cannot run on a laptop, it does not belong in the pipeline.
- Tag images with the commit hash.
- Make every deploy reversible in one command.
- Alert on symptoms, not causes.
- Review infrastructure plans like code.
The pipeline, step by step
We watch four signals for every service: request rate, error rate, latency at the 95th percentile, and saturation. Anything else goes on a secondary dashboard that we look at when those four disagree.
Premature optimization is the root of all evil.
Donald Knuth, Structured Programming with go to Statements, 1974
Quick takeaway
Small, reversible, observable. In that order.
What we monitor
Small releases
Ship many tiny changes instead of one big one.
Easy rollback
Every deploy can be undone in a minute.
Four signals
Rate, errors, latency, saturation.
Infrastructure changes go through the same review as application code. Terraform plans are posted to the pull request so reviewers can see exactly what will change before anything is applied.
If you try this on your own project, start with one page or one service rather than the whole codebase. A small experiment gives you real numbers to argue with, and it is easy to roll back if the idea does not fit.
We have been running this approach for several months now. The initial setup took about a day, and the ongoing cost is close to zero once the team knows where to look.
Wrapping up
As always, the examples are simplified to fit on a page. The production version has more error handling and logging, but the shape of the code is the same.
We run this in production at a fairly large scale and can confirm the caveats are real.
A repo is on the way. I will link it in the post once it is cleaned up.
We run this in production at a fairly large scale and can confirm the caveats are real.