Hosting from a boat using Starlink

So I've been living on a 34-foot sloop (actually a cutter rig, technically, but who's counting) for about eight months now and figured I'd share how the hosting situation shakes out when your datacenter is literally floating. Back in my day (2019) we thought 4G was revolutionary for remote work, but Starlink changed the game (though not without caveats (actually significant ones (latency being the big killer)))).

My setup:

• Raspberry Pi 4 (8GB, running Alpine Linux 3.19.1 actually) in a waterproof enclosure
• UPS
• Starlink Mini

The dish is gimbal-mounted (DIY, based on Open astro tracker v1.13 firmware) to compensate for wave motion. Key finding: uptime correlates strongly with sea state. Calm days (under 1m swell) I see 15-20ms to shore stations, but in 3m+ seas the dish can't maintain lock and you're looking at 40-60% packet lo

Electricity costs on a boat with solar, now that's optimization. My rigs pull 2.4kW continuous, can't imagine running that off panels. The starlink latency to shore stations—have you measured how it affects your actual workload profitability? I find anything over 30ms starts cutting into margins. Marine setup's interesting though, redundant power is something I should look into for my secondary location.

Grabs popcorn

Following 🍿

Oh man this takes me back to when I tried to host a mumble server (remember mumble? (the voice chat thing (before discord killed it))) from a literal tent in 2020 which is not a boat obviously but the principle of hostile environments and networking challenges feels spiritually related and I had this whole thing with a yagi antenna I built from pvc pipe and coat hangers (seriously, coat hangers, the wire is close enough to the right gauge if you straighten them carefully) pointing at a cell tower six miles away through trees that swayed in the wind which created this periodic fading pattern every time the wind gusted and I thought about building a mechanical tracker but then realized I could just use two antennas in diversity configuration which worked okay but not great and the real problem was condensation inside the enclosure (same as your salt issue probably (different mechanism (hum

(Kidding, obviously you can't, FCC regulations and all that.)

More seriously, your Pi setup is solid but consider adding a reverse proxy on shore somewhere. I run nginx (docker compose, obviously) on a €3 VPS at Hetzner that tunnels back over wireguard. When the boat link flaps, at least your services stay reachable. The shore proxy handles TLS termination too, which saves the Pi's CPU for actual workloads.

For antenna stabilization, have you looked at open-source telescope mounts? INDI drivers (libindi 2.0+) support a lot of commercial hardware. Docker compose stack: indi-server, your tracking algorithm, mqtt for telemetry. I self-host all of it, obviously. The graphs don't lie—my uptime went from 92% to 99.7% after adding shore-side failover.