MiniStore
A production-style object storage service — a small internal S3 — built phase by phase to learn infrastructure, observability and disaster recovery.
On this page
- Ended
- Ongoing
- Type
- project
Built with
Overview
A production-style file and object storage service — a small internal S3 — built phase by phase specifically to practice backend infrastructure, observability, failure handling and disaster recovery, rather than to ship a product.
Core rule: file bytes live in MinIO; metadata (owner, name, size, version) lives in PostgreSQL. Everything else in the design follows from keeping those two stores cleanly separated.
How it works
Built in explicit phases, each with its own doc covering what's being built and why before the code, and what was actually learned after:
- Phase 0 — overview and architecture: done
- Phase 1 — object storage + MinIO in Docker: done
- Phase 2 — backend integration (Express + Postgres + MinIO): done
- Phase 3 — secure file uploads: next
- Phases 4–15 — the rest of the roadmap, covering load testing (k6), monitoring (Prometheus + Grafana) and CI (GitHub Actions)
Stack: Node.js + Express + TypeScript, PostgreSQL via Prisma, MinIO, JWT, Docker Compose, Jest/Supertest for tests.
Architecture
One rule shapes everything: file bytes live in MinIO, metadata lives in PostgreSQL. Keeping them apart is what makes versioning, access control and recovery tractable — and what makes the failure modes interesting enough to be worth practising on.