Field telemetry reference architecture — homestead + 120-acre farm¶

Purpose¶

Provide a single first-draft reference for how sensors, gateways, uplink, broker, and operator interfaces could fit together across SITE_HOMESTEAD and SITE_FARM—without claiming this is your deployed stack. Use it to spot single points of failure (SPOFs), missing boundaries, and where truth lives for events.

Package hub (5 ac + 120 ac framing, SoR, security): Reference architecture — 5-acre home base + 120-acre farm.

Demory SITE_FARM sensor stack (long-range RF, gateways, Starlink as WAN only): Farm sensor architecture — Demory farm site · Sensor checklist matrix — Demory farm.

Integration / package boundary: New RF/sensor/vendor ingests activate into the relevant cluster source-notes (Evidence summary) and the Demory pages above—not into this reference as a dump. Platform (broker, TSDB, k3s observability) routes through Platform doctrine package — homelab / farm edge and Telemetry system of record — boundaries and authority. WAN/power scheduling ties to Off-grid power and field networking hub and backup/DR when telemetry affects restore order.

Scope¶

In scope: Logical blocks, data paths, assumptions, and questions for a split-site, partly instrumented farm.
Out of scope: Vendor selection, exact SKUs, network security hardening (link to future cybersecurity hub); regulatory compliance.

Assumptions (edit freely)¶

Placeholder	Meaning
`SITE_HOMESTEAD`	Smaller site; may host home LAN, Home Assistant or equivalent, internet uplink.
`SITE_FARM`	Larger parcel; may have weak broadband, solar/battery field power, LoRa/Meshtastic/LoRaWAN at edges.
`BROKER_LABEL`	MQTT broker (self-hosted or cloud)—TBD.
`SOR_TELEMETRY`	System that is authoritative for long-term sensor history—TBD (TSDB, DB, vendor cloud).

Roles: Home Assistant, farmOS, MQTT, TSDB (draft binding)¶

Layer	Typical role	Conflicts when…
MQTT broker	Transport and pub/sub between devices and services	Retained vs live state misunderstood
Home Assistant	Automation, dashboards, consumer device hub at homestead	Treated as legal record for tax or certification without export
farmOS	Structured assets, logs, geometry (when used)	Out of sync with field reality if not updated
TSDB / PostgreSQL	Durable telemetry history / joins to map	Duplicate writes from HA and direct ingest—need SoR rules (see backlog P0 #6).

farmOS model refs: farmOS Assets, farmOS Logs.

Connects the strands¶

Strand	Connection
C	Radio choice and gateway SPOFs—`LoRaWAN vs Meshtastic`, `Fixed gateway tower vs mobile`.
D	farmOS + HA + DB—`farmOS vs lightweight`, `Data storage`.
G	Time for correlated logs—`Time synchronization`.

Reference diagram (logical, text)¶

text [ Field devices @ SITE_FARM ] e.g. soil moisture, tank level, gate sensor, camera (placeholders) | radio path (LPWAN / Wi‑Fi / mesh — TBD) v [ Edge gateway / concentrator @ SITE_FARM ] power: `FIELD_POWER_ASSUMPTION` | backhaul (cell / fiber / point-to-point / mesh — TBD) v [ IP network ] --------------------> [ BROKER_LABEL ] <---- subscriptions --- [ Operator UI / HA / phone ] | ^ | | [ SITE_HOMESTEAD LAN ] -------------------------+ (VPN or same LAN if bridged — TBD)

Not drawn: time sync (NTP/PTP/GNSS)—see Time synchronization — NTP and PTP sources.

Key questions¶

If SITE_FARM uplink dies, what is still true locally? (Cached automations? Gate state unknown?)
Who is authoritative for “tank is low”—broker last message, field node, or a database deduplicated state?
Where do alerts originate—broker rule engine, Home Assistant, external SaaS? (Avoid three alert channels with no priority.)
Power: Which edges are battery + solar vs grid? What happens after BATTERY_DAYS_AUTONOMY cloudy stretch?
Admin access: How do you reach gateways for updates without exposing everything to the public internet?

SPOF checklist (fill in)¶

Component	If it fails…	Mitigation idea (placeholder)
Internet at `SITE_HOMESTEAD`	You may lose remote view of `SITE_FARM`	Local alerts on-site; degraded mode SOP
Broker `BROKER_LABEL`	All MQTT consumers blind	HA queue / bridge strategy; broker redundancy—TBD
Field gateway	Subtree of sensors dark	Spare gateway? Radio path diversity?
Time sync	Correlated logs wrong	GNSS/NTP strategy per `Timing on the farm — synthesis`

Decision criteria¶

Choice	Favor when	Avoid when
Concentrate gateways at one mast	Clear line-of-sight, simpler power	Single lightning/wind failure domain
Distributed gateways	Terrain blocks RF; need mesh	More firmware surfaces to patch
Cloud broker	No ops capacity	Latency, privacy, outage sensitivity

Data storage (farm and edge stacks) — PostgreSQL/MQTT docs entry.
LoRa, MQTT, and gateway bridges — patterns and sources.
Field network IoT comparisons (HaLow, LoRa, NB-IoT) — radio tradeoffs.
Farm data, farmOS, and agriculture lab builds — records vs telemetry.
Telemetry system of record — boundaries and authority — authority matrix for records vs telemetry vs dashboards.
Spatial data and farm asset registry standard — bind device IDs to places.
Automation degraded modes and manual fallback SOP — behavior when this architecture breaks.

Open items¶

Replace placeholders with one-page network list: subnets, VPN, broker hostnames (even if private).
Add ASCII or exported diagram when stable.
Align admin paths with Remote access and operational security model.

First draft: logical only; validate against a real diagram and an electrician/radio path survey before relying on it for safety.