← The Volumes

Volume IV · Infrastructure · Reliability

The Quiet Infrastructure

Two servers, roughly thirty production systems, and a backup pipeline nobody has to remember. The discipline of operating what you build.

Problem
Around 30 production projects on two servers, with backups that depended on someone remembering.
Outcome
Automated offsite PostgreSQL backups to Backblaze B2, pruned after three days, on a hardened Docker and Nginx setup.
Diagram: Around 30 production projects on two servers, with backups that depended on someone remembering.
Principal
Zylo Technologies
My role
Engineer of record for production infrastructure
Period
2025 — Present
Tech used
Linux · Docker · Nginx · PostgreSQL · Redis · Backblaze B2 · CI/CD

Chapter 1 · The problemThe Thesis

Most servers do not fail loudly. They fall over quietly under real traffic because nobody set them up to handle it. Reliability is not a feature you add; it is the absence of a thousand small omissions.

Chapter 2 · The client & my roleThe Catalyst

At Zylo, the products I build also have to run somewhere. I manage the production estate: two servers carrying around thirty projects, on Linux with Docker, Nginx, Redis, and PostgreSQL.

One engineer owning the whole chain, from the mobile screen to the process on the server, removes the gap where most production failures hide.

Chapter 3 · The hard partThe Friction

Thirty projects on two machines means one bad night can take down many clients at once. Backups that depend on a person remembering are not backups.

The decision was automation with a retention discipline: compressed PostgreSQL backups, shipped offsite to Backblaze B2, with old copies deleted after three days, so storage never becomes the reason the next backup fails.

“A backup you have to remember is a backup you will forget.”

Chapter 4 · The resultThe Monument

The infrastructure runs without ceremony. Data leaves the building every cycle, retention prunes itself, and production is something the team stops worrying about.

2
Production servers under management
~30
Projects running in production
3 days
Self-pruning offsite backup retention