Case study
Trackflow Ops
Incident command centre
Restore testedPostgreSQL and Storage backups
Live demo
Hosted on my own server: the first load after a quiet period can take a little longer.
What it is and what it solves
An incident command centre: detection, automatic incidents, live comments, on-call rotations, status pages and a Chaos lab. It solves the gap between an alert firing and a team coordinating around it, with every change reaching everyone in real time. The Chaos lab injects errors, latency or outages into demo services so you can watch an incident open and close.
Business version: TrackFlow Maintenance →
Architecture
- Browser → nginx
- nginx → API
- API → PostgreSQL
- Worker → PostgreSQL
- API → Kafka
- Kafka → Worker
- API → Redis
- Prometheus · Grafana → Worker
Simplified from the project stack; the repository is private.
Three technical decisions
Row-level security in PostgreSQL
Access rules live in the database (Supabase RLS), so the API, the worker and any future client share the same authorisation instead of re-implementing it.
Kafka with a transactional outbox
Incident changes and their events are committed together, so detections, notifications and timelines stay consistent with the stored state.
Observability as part of the product
Prometheus and Grafana watch the platform itself, and n8n handles integrations, so the incident tool is not blind to its own failures.
Tests
- CI runs the build and tests on every change and is green.
- Backup restore tested for both PostgreSQL data and Storage files.
- Kafka health probe passing.
Verified
- CI green.
- Backup restore exercised end to end.
Not claimed
- Figures come from test runs, not from production traffic.
Stack
Code and CI
Private repository · code walkthrough on request