← All work Fleet monitoring

Hashfield

Monitoring, inventory and safe control for Bitcoin mining fleets at remote sites.

Kind
Fleet monitoring
Stack
Go · Rust · React
Parts
Site agent, web app, public site
Site
hashfield.org
The Hashfield fleet overview with hashrate, power and efficiency across six sites, light theme
The Hashfield fleet overview with hashrate, power and efficiency across six sites, dark theme
LightDark View larger
The fleet: hashrate, power and efficiency across six sites.

Overview

Hashfield watches ASIC miners at sites nobody can drive to every day. A small agent at each site reads every miner every minute and keeps writing to disk when the cell link drops, then catches up in order when it's back, with no gaps and no duplicates.

From a summary of every site down to one hashboard's chip temperatures, it shows a failing board days before it drops, follows every part by serial across the miners it has been in, and rolls changes out in waves.

What we built

Site agent

Reads every miner every minute and keeps collecting through outages, then backfills in order.

Early warning

Per-board hashrate and chip temperatures show a failing hashboard before it fails.

Inventory

Containers, racks and slots as they really are, with QR labels scanned from a phone.

Parts by serial

A hashboard keeps its history across every miner it has been in.

Safe control

Every action shows what it will touch, asks before big ones, and rolls out in waves.

Alerts

Racks in maintenance hold their alerts until the work is done.

Screens

The miners list with status, site, rack slot, hashrate and chip temperatureThe miners list with status, site, rack slot, hashrate and chip temperature
Every miner, offline first, with its rack slot and chip temperature.
The fleet heatmap: every miner as a row over a weekThe fleet heatmap: every miner as a row over a week
A week of every miner, colored by hashrate against expected, worst first.
A miner's page with its hashrate by board as one board drops outA miner's page with its hashrate by board as one board drops out
One miner's boards over eleven days, as one fades and drops out.
A site's page with miners online, hashrate, power and an alertA site's page with miners online, hashrate, power and an alert
A site: miners online, hashrate, power and coverage.
A site's container shown as racks and positions with miners in their slotsA site's container shown as racks and positions with miners in their slots
A container's racks, with each miner in its slot.
Active alerts by severity, with acknowledged and held alertsActive alerts by severity, with acknowledged and held alerts
Alerts by severity, one held for maintenance.
A site agent's sync health showing every record accounted for after an outageA site agent's sync health showing every record accounted for after an outage
A site agent's sync health after an outage it backfilled.
A hashboard's page with its readings across two minersA hashboard's page with its readings across two miners
A hashboard followed across the two miners it has been in.