orionkabeza/disaster-warning-map

Live natural-hazard map for Africa that plots events from NASA EONET and USGS, then finds the nearest hospitals and clinics from OpenStreetMap. Deployed across two web servers behind a HAProxy load balancer. # East Africa Disaster Early-Warning Map ALU Web Infrastructure summative — "Playing Around with APIs". ## Demo video A short (~2 min) walkthrough showing the app running locally and through the load balancer: **→youtu.be ## Live deployment **→ patient0.tech** Served over HTTPS (Let's Encrypt) through a **HAProxy** load balancer (`Lb01`) doing round-robin across two identical web nodes: | Role | Host | Serves | |---|---|---| | Lb01 | HAProxy, `www.patient0.tech` | round-robin + TLS termination + HTTP→HTTPS redirect | | Web01 | `7090-web-01` | nginx (static + `/api` reverse proxy) → Express on `:3000` | | Web02 | `7090-web-02` | identical to Web01 | | DB | Neon Postgres (`eu-central-1`) | users + saved settings, over TLS | Round-robin is verifiable live — each response names the backend that served it: ```bash for i in $(seq 10); do curl -s patient0.tech; echo; done # → alternates {"ok":true,"host":"7090-web-01"} / {"ok":true,"host":"7090-web-02"} ``` The load balancer uses an **application-level health check** (`option httpchk GET /api/health`), so a node whose Express backend is down — even with nginx still up — is pulled from rotation rather than returning 502s. The exact working LB config is committed at `deploy/haproxy/haproxy.cfg`; full steps are in `deploy/RUNBOOK.md`. ## The problem Public hazard feeds (earthquakes, floods, storms) tell you *that* something is happening, but not *what to do about it*. This app plots natural-hazard events across Africa from **NASA EONET** and the **USGS earthquake feed**, and layers on the nearest health facilities from **OpenStreetMap** (via the Overpass API), so a viewer near a flagged event can answer "where do I go" in the same view, not by cross-referencing three separate apps. It is explicitly not a weather app, joke generator, or converter — it combines three independent live data sources into one decision-relevant view, with filtering, sorting, and p …