🟢 aswincloud status

A self-hosted, BetterStack-style live status page for a home network.

Built on the Cloudflare edge so the page stays up even when home — or the prober — is down.

▶ Live demo: status.aswincloud.com

---

How it works

A status page that lives at home dies with home — useless. So the page runs on Cloudflare's global edge (a Worker + a D1 SQLite database), and an external always-on server runs a tiny Docker prober that checks your home and pushes heartbeats in.

 ┌─ external always-on server ─┐         ┌──────────── Cloudflare edge ───────────┐
 │  prober (Docker)            │         │  Worker  ·  D1  ·  per-minute cron      │
 │   every 30s:                │  HTTPS  │                                         │
 │     check home  ───────────────────▶  │  POST /api/ingest   → record + detect   │
 │     POST heartbeat          │ (Bearer)│                       up/down flips     │
 └─────────────────────────────┘         │  GET  /api/status   → JSON for the page │
                                         │  scheduled()        → watchdog + prune  │
   visitor ─── GET / ─────────────────▶  │  static assets      → the status page   │
                                         └─────────────────────────────────────────┘
                                                        │ on a down/up flip
                                                        ▼
                                            📧 Email · 💬 Slack · ✈ Telegram

Both failure modes surface correctly:

What breaks	How it's caught
🏠 Home is down	The prober's check fails → pushes down → recorded, incident opened, alerts fire.
🖥️ The prober/server itself dies	No heartbeats arrive → the Worker's per-minute cron watchdog sees stale data (>2 min) and records a synthetic outage. No silent green.

A note on the check type. The demo monitors torrent.aswincloud.com, which is proxied through Cloudflare. A ping to it would only reach Cloudflare's edge (always up) and show false-green — so the monitor uses an HTTP check that flows through Cloudflare to the home origin: 200 when home is up, 52x when it isn't. For a direct (non-proxied) host, ping or tcp work too. The prober supports all three.

---

✨ Features

• Overall banner — All systems operational · Partial outage · Major outage
• Per-monitor cards with a live status pill
• 90-day uptime bar — hover any day for that day's uptime %
• Uptime % over 24h / 7d / 30d
• Response-time sparkline (last 2 hours, SVG, breaks the line on downtime)
• Incident timeline — ongoing + resolved, with durations
• Alerts on every down/recovery — Email (Resend), Slack, and Telegram, each independent
• Dark / light theme toggle · fully responsive · zero frontend dependencies
• Config-driven — add a monitor by editing one JSON file; the page auto-discovers it

Supported check types: http · tcp · ping.

---

🗂️ Layout

wrangler.jsonc     Worker + D1 + cron + static-assets + custom-domain config
schema.sql         D1 tables + starter monitor seed
src/
  index.ts         fetch() router + scheduled() cron watchdog
  api.ts           /api/ingest (auth) + /api/status (edge-cached JSON)
  db.ts            D1 helpers + Env types + tunables (STALE_MS, RETENTION_MS)
  stats.ts         uptime %, 90-day buckets, latency series, incidents
  alerts.ts        email / Slack / Telegram — each activates only when its secrets exist
public/            The status page — index.html · styles.css · app.js
prober/            The Docker prober (copied to the external server)

---

🚀 Part 1 — Deploy the page (Cloudflare)

Requires a Cloudflare account. For the custom domain, your zone should be on Cloudflare.

npm install

# 1. Create the D1 database, then paste the printed database_id into wrangler.jsonc
npx wrangler d1 create status-db

# 2. Create the tables (remote = the real database the Worker uses)
npx wrangler d1 execute status-db --file schema.sql --remote

# 3. Set the shared ingest secret — SAVE it, the prober needs the same value
#       openssl rand -hex 32
npx wrangler secret put INGEST_TOKEN

# 4. Deploy
npx wrangler deploy

Custom domain → in wrangler.jsonc the routes block maps status.aswincloud.com; since the zone is on Cloudflare, DNS + TLS are provisioned automatically on deploy. Before a domain is attached, the Worker is live at home-status.<account>.workers.dev.

---

🐳 Part 2 — Run the prober (external, always-on server)

Copy the prober/ folder to that server, then:

cd prober
cp .env.example .env
nano .env                       # INGEST_TOKEN = the same value from step 3 above

# config.json already points at https://status.aswincloud.com/api/ingest
docker compose up -d
docker compose logs -f          # expect: "ok — home-network:140ms"

restart: unless-stopped + Docker-on-boot keep it running across reboots and crashes.

---

✅ Verify end-to-end

1. Open the page — the monitor turns green within ~30s; sparkline + 90-day bar fill in.
2. curl -s https://status.aswincloud.com/api/status | jq '.overall, .monitors[0].up' → "operational", true.
3. Outage test: docker compose stop the prober. Within ~2 min the card flips red, an incident opens, alerts fire. docker compose start → it recovers and the incident shows resolved with a duration.

---

➕ Add another monitor

Edit prober/config.json, then docker compose restart. The new card appears automatically.

{
  "monitors": [
    { "id": "home-network", "name": "Home Network", "type": "http", "target": "https://torrent.aswincloud.com" },
    { "id": "jellyfin",     "name": "Jellyfin",     "type": "tcp",  "target": "192.168.1.50:8096" },
    { "id": "router",       "name": "Router",       "type": "ping", "target": "192.168.1.1" }
  ]
}

ping/tcp to LAN addresses require the prober to have a network route to them. To delete a monitor's history immediately: npx wrangler d1 execute status-db --remote --command "DELETE FROM monitors WHERE id='jellyfin'; DELETE FROM checks WHERE monitor_id='jellyfin';"

---

🔔 Alerts

src/alerts.ts fires on every down/recovery transition. Each channel activates only when its secrets are present — run any combination, or none. Secrets live in Cloudflare, never in this repo.

📧 Email (Resend)

npx wrangler secret put RESEND_API_KEY   # resend.com — sending domain must be verified
npx wrangler secret put ALERT_FROM       # "aswincloud status <status@aswincloud.com>"
npx wrangler secret put ALERT_TO         # aswin@aswincloud.com  (comma-separated for several)

💬 Slack — create an app with the chat:write scope, /invite the bot to a channel, grab the channel ID (C0…):

npx wrangler secret put SLACK_BOT_TOKEN
npx wrangler secret put SLACK_CHANNEL

✈ Telegram — create a bot via @BotFather, read your chat id from https://api.telegram.org/bot<TOKEN>/getUpdates:

npx wrangler secret put TELEGRAM_BOT_TOKEN
npx wrangler secret put TELEGRAM_CHAT_ID

After setting secrets, just push to main (or npx wrangler deploy) — no code change.

---

🔄 Continuous deployment

Connected to Cloudflare Workers Builds: every push to main runs npx wrangler deploy; pushes to other branches run npx wrangler versions upload (preview). Worker secrets persist across builds.

The prober is not part of this pipeline — it's a container on an external box. Update it there with git pull && docker compose up -d --build.

---

🧪 Local development

npx wrangler dev                                            # local Worker + D1 at :8787
npx wrangler d1 execute status-db --file schema.sql --local # seed the local DB
cd prober && INGEST_TOKEN=devtoken INGEST_URL=http://localhost:8787/api/ingest node prober.js
# (give wrangler dev the matching token via a .dev.vars file: INGEST_TOKEN="devtoken")

Open http://localhost:8787.

---

📝 Notes & limits

• Fits comfortably in Cloudflare's free tier (Workers + D1 + cron).
• Raw checks are pruned after 90 days; uptime % and the day-bar use indexed aggregates, and /api/status is edge-cached ~15s so page loads never re-scan.
• Tunables: STALE_MS (2 min) and RETENTION_MS (90 d) in src/db.ts; check interval in prober/config.json (intervalSeconds).
• Secrets (INGEST_TOKEN, alert credentials) are Cloudflare Worker secrets — never committed.

---

_{Built with Cloudflare Workers · D1 · Docker — and zero frontend dependencies.}