Infrastructure
We run our own: the servers, the pipeline that ships every change, and the alerts that tell us when something's wrong. It's what lets a small team ship often and still sleep.
How a change ships
Run a release, or send a broken one and watch the checks stop it before production.
Merge
A change lands on the main line once its checks pass.
Build
One image is built from that exact commit, and only that image ships.
Staging
It runs on staging first, a copy of production.
Checks
Health checks and smoke tests must pass before it goes any further.
Production
It takes over from the old version while connections drain, so nobody is cut off.
Watched
Metrics and alerts watch it, and a bad release is rolled back.
Every stage passed: the release is live, and nobody was cut off.
What we run
Self-hosted servers
Set up from code, so any one of them can be rebuilt from scratch.
Staged releases
Each change is built once, tried on staging and promoted only when its checks pass.
Zero-downtime releases
The new version takes over while the old one drains, so nobody is cut off.
Observability
Metrics, logs and traces, with alerts that reach us on our phones.
Backups
Taken every night, and each one restored and checked.
What runs on it
KOL Blaster, CampusGains and this site, each released the same way.
Next: how we work
A small studio, end to end.