What FearMiner collects

what

22 fields in 12 groups, the ones the egress line names at start: event, install, format, version, system, mining, hardware, hashrate, features, uptime, errors, pool. The hashrate goes as a wide bucket, never the figure.

never

the wallet, the worker name, a login or a password, the machine's name, the IP address, a serial number, a path, a configuration value, anything of the cockpit. The full list is below.

when

once the run is under way, every 15 minutes, and at a clean stop; never more often than every 5 minutes.

where

telemetry.fearminer.com, a service of its own, walled off from the cockpit and its relay. The country is kept, never the address.

kept

raw points about 3 months; daily counts without any install id, longer.

off

--no-telemetry, or FEARMINER_NO_TELEMETRY=1: no install id is drawn or kept, nothing is sent, and the miner loses no feature.

see it

fearminer telemetry show prints the beat this machine would send, and sends nothing.

One beat, as it leaves the rig:

{
  "schema": 1,
  "kind": "beat",
  "first": false,
  "previous_version": null,
  "install": "5f0c3b8e2a9d4c71b6e8f0a1d2c3b4a5",
  "interval": 15,
  "version": "1.5.1",
  "build": "f6f9d649c",
  "os": "linux",
  "os_release": "debian-13",
  "mining_os": null,
  "algo": "quantus",
  "engine": "1.0.1",
  "engine_build": "linux-x86_64-cuda",
  "driver": {"nvidia": 575, "cuda": 12},
  "gpus": [{"model": "GeForce RTX 4070 Ti", "count": 2}],
  "cpu": {"vendor": "amd", "model": "Ryzen 9 9950X", "cores": 16},
  "rate": {"bucket": "1G-2G", "unit": "H/s"},
  "features": ["doh", "failover", "helper", "service", "tls"],
  "uptime": "lt1d",
  "errors": {"E116": 3},
  "pool_host": "pool.example.com"
}

The ingestion checks each report against the schema, adds the country and writes one data point; what is read from the points is counts, such as how many installs run each version, system or algorithm. The last column says what becomes of each field.

The event event #

Which moment the report is: a start, a beat or a stop.

The install id install #

A random number that stands for this installation and for nothing else.

The format format #

How to read the rest of the report.

Version and build version #

Which release of the miner runs.

The system system #

Where the miner runs.

What is mined mining #

The algorithm, the engine that mines it, and the drivers' major versions.

The hardware hardware #

Models and counts; never a serial number or a card's identifier.

The hashrate, as a bucket hashrate #

A wide bucket, never the figure; the buckets are below.

Features in use features #

Which settings are in use, as named flags, never their values; the flags are below.

Uptime, as a bucket uptime #

How long this run has lasted, never to the second.

Error codes errors #

The stable codes the miner printed since the last report that got an answer, with their counts; never a message.

The pool's host pool #

Which pools the rigs mine on.

0 is a rig at zero and 0-1 a rig under 1; every other bucket runs from one value of the series to the next, 1, 2, 5, 10, 20, 50, 100 and so on, written with k, M, G, T and P and no decimals. A few rates, and what is sent for each:

Wide enough to tell a rig's size, and a version that got slower on the same hardware; too wide to give a farm's exact figure.

• wallet
• worker name
• pool login or password
• machine name
• MAC address
• card or disk serial number
• GPU UUIDs or PCI bus ids
• file paths
• configuration contents or values
• IP address (the ingestion reads the country from Cloudflare and never the address)
• anything of the cockpit or relay (rig fingerprint, keys, codes)

The miner's tests hold every report to the schema: a field outside it fails the build, and so does a test configuration's wallet, worker, password, host name or path found anywhere in it.

• A crash sends nothing: the missing beat says it.
• A report that does not go through is lost: never queued, never sent again, never sent later in a burst. Only its error counts wait for the next report.
• The miner never reads the answer and logs a failure at debug level only: it mines the same whether telemetry.fearminer.com answers, refuses or is down for weeks.
• In the figures, a rig connected now is a rig seen in the last 20 minutes, so at the default interval one late beat does not drop it.

Why not every minute: a rig does not change its algorithm, its cards or its country from one minute to the next, so a minute cadence would multiply the volume by fifteen and tell nothing more. Following one rig minute by minute is what the cockpit does for its owner, not what an anonymous count is for.

• Nothing of the cockpit is ever sent here: no rig fingerprint, no fleet or rig key, no enrolment code, no relay address. The cockpit flag says a rig is enrolled, never where.
• The install id is drawn at random and derived from nothing: not from the rig's name in the cockpit, not from a key, a hash or a date.
• The two are never joined: no table, query, export or support ticket puts a rig of the cockpit next to an install of the telemetry.
• The relay is never a source of telemetry: what it knows about the rigs connected to it stays in it.

The reason is structural. The cockpit promises that its relay holds nothing it can read. A relay that could tell that one of its rigs is an install seen in some country with six cards and some hashrate would hold exactly what it promised not to hold. So the two stay apart, even when their counts of rigs differ.

Stated plainly: a hashrate bucket, GPU models and a country, taken together, can point at a large farm, because there are not ten installs of 400 cards of one model in one country. That is why the hashrate is a wide bucket, why the location stops at the country, and why the install id carries nothing else. It narrows the risk; for the largest farms it does not remove it.

A farm that would rather send nothing at all uses --no-telemetry. It turns everything off: no install id is drawn or kept, no start, no beat, no stop, no error count; and the miner loses no feature: mining, the cockpit, the API and the notifications work the same.

$ fearminer -o stratum+ssl://POOL:PORT -u YOUR_WALLET --no-telemetry
$ FEARMINER_NO_TELEMETRY=1 fearminer -o stratum+ssl://POOL:PORT -u YOUR_WALLET
# fearminer.toml
[output]
no_telemetry = true

On by default, because only representative numbers are worth collecting: an opt-in would lose most reports and hide every less common system, backend or error. The price of that choice is this page, the egress line and one switch that turns it all off.

1. 2026-09-25 schema 1 The flag tls-insecure is added, still schema 1: that the pools' certificates are not verified, never which pool.
2. 2026-09-24 schema 1 First version of the telemetry and of this page, schema 1.