The tablet that saved my on‑call nights (and the things that still break)
How I turned an old Android tablet into a wall-mounted on‑call console with Grafana, Uptime Kuma, Tailscale and a local runbook — and the real tradeoffs I ran into.
Written by: Arjun Malhotra
It was 3:07 a.m. The pager buzzed. My laptop was asleep under a charging cable. The router had blinked out once earlier in the night during load‑shedding. I fumbled for my phone, opened the VPN app, logged in, then waited for a dashboard to load over my flaky home 4G hotspot. Ten minutes later the alert was acknowledged, but the database was still degraded and I’d already lost a chunk of sleep.
That morning I decided to stop depending on a single, personal device for on‑call. I dug out a Redmi tablet I’d bought second‑hand for ₹3,200, wall‑mounted it next to my desk, and made it a one‑job device: status, metrics, runbooks.
This is what I built, why it works in India, and the small failures you should expect.
What I actually put on the wall (hardware + software, with costs)
- Tablet: ₹3,200 (second‑hand 8” Android). You can use a newer cheap tablet — I’ve seen ₹6,000 options that work fine.
- Wall mount and USB cable: ₹450.
- Power: standard USB charger (5W). If you want battery backup add a ₹1,200 10,000 mAh powerbank.
- Software:
- Fully Kiosk Browser (free tier) to lock the device into a URL and keep the screen awake.
- Grafana hosted on a small VPS (my monitoring stack sits on a ₹300/month VPS).
- Uptime Kuma for ping checks (self‑hosted).
- Tailscale on the server and Android to avoid exposing dashboards publicly.
- A tiny local static HTML runbook stored in Fully’s local web content (quick checklists and SSH jump host addresses).
Total outlay: under ₹4,000 if you re‑use an old tablet. Monthly add-on: the VPS ~₹300.
Why this combo? Fully Kiosk locks the tablet to the dashboard, keeps the screen on, and supports a whitelist of local contents. Tailscale avoids hair‑pulling NAT/sketchy exposed ports. Grafana gives me visual state. Uptime Kuma gives me immediate ping status and can be configured to open the runbook URL.
Why this is actually useful in an Indian setup
- Power cuts are routine. A wall‑mounted tablet on an always‑on UPS or a small powerbank survives short outages that take my laptop offline.
- Mobile data is expensive and slow. Tailscale + small VPS means the tablet doesn’t re‑download large assets; it just renders lightweight dashboards and runs a local cached fallback.
- Office/home internet flakiness: when Wi‑Fi drops, the tablet can still show the last successful snapshot of metrics and a cached runbook to follow step‑by‑step.
- Visibility: family members (and sometimes roommates) see that I’m on‑call and avoid unplugging things. That small social win matters.
I can acknowledge a pager from my phone, but I don’t need to faff with the laptop to run the first checks. I can read CPU heatmaps, check error rates, and follow the exact runbook for common failures — without logging into anything.
The honest failures and tradeoffs
This isn’t magic. I ran into a few ugly realities.
- Push notifications are unreliable on a kiosk device. Android aggressive battery killing and Doze mode meant I sometimes didn’t get an alert if the device wasn’t actively rendering. Fully keeps the screen awake, which mitigates this, but it also burns power and shortens tablet life. I ended up relying on Uptime Kuma’s loud alarms (hosted on the VPS) rather than Android push.
- Screen on = burn‑in risk and shorter hardware life. The cheap tablet’s backlight will dim after a year of always‑on use. I disabled high‑brightness night mode and used a screen saver that shows an animated logo to minimise static content.
- Physical mounting failed once. The Command strip I used came off after a humid week and the tablet fell. The glass survived, software didn’t. I now use a cheap screw mount and a micro‑USB cable anchor. Safe mounting costs another ₹200 but is worth it.
- Offline troubleshooting is still hard. The tablet can host a runbook and SSH jump addresses, but I still need a full keyboard and terminal for deep fixes. The tablet is a triage console, not a replacement laptop.
- When the VPS I used for dashboards went down (my bad — I’d co‑hosted monitoring and a small test webapp together), the tablet showed nothing useful. I introduced redundancy: small backups, a second uptime ping from a different region, and a static cached page that lists emergency contacts and a minimal checklist.
One constraint that genuinely changed how I use the device: I stopped treating it as a source of live logs. Instead I design dashboards around what I can do with just the tablet — service health, top 3 incident causes, and links to precise runbook steps. Anything requiring logs or rebuilds, I leave for the laptop.
Day‑to‑day habits that made it reliable
- Keep a local cached runbook HTML in Fully Kiosk. If the network is gone I can still follow steps.
- Use short auto‑refresh intervals (30–60s) for Grafana panels, and a small nginx cache on the VPS to reduce payload.
- Automate a loud alarm in Uptime Kuma that calls my phone via HTTP → third‑party call service; it’s louder and harder to ignore than a push.
- Rotate the tablet weekly: quick reboot and visual inspection. Cheap hardware accumulates quirks.
- Put a sticky label with SSH jumphost and my Tailscale key name on the mount. Old school, but it saved me twice.
What I walked away with: a wall tablet doesn’t replace your laptop — it reduces the number of times you need to fire up a laptop at 3 a.m. It makes the first 10 minutes of an incident structured and calmer, and that alone preserves sleep and judgment. The tradeoffs are physical: battery life, occasional mounting drama, and handling Android’s background quirks. If you want fewer late‑night panics and you can spare a ₹3–6k tablet, it’s a tiny investment with outsized returns.
I still don’t get every alert on the tablet. I still log into my laptop for hard fixes. But I no longer waste ten minutes getting attached to a flaky network before I start triaging. That small friction removed has saved me more sleep than any monitoring dashboard ever promised.