Smart technology architecture audit (strategic audit)¶
Purpose: Dedicated architecture audit for the smart-technology slice of a future two-site operation (~homestead + ~120-acre farm), derived from Strategic audit — decision-safe operations for a two-site smart farm and Implementation backlog — strategic audit (P0–P3). This page names what exists in the wiki, what is missing, where the architecture is fragmented, and what reference pages, SOPs, and diagrams are still required.
Out of scope here: Pure business / revenue economics (see Business viability and farm economics — gap analysis); biological stocking conclusions.
1. What the wiki already covers (by focus area)¶
| Focus area | What exists (representative pages) | Maturity note |
|---|---|---|
| Field telemetry | Field telemetry reference architecture — homestead + 120-acre farm (draft); Tracking chickens — motion sensors over LoRa and MQTT |
Draft logical stack only; no deployed topology |
| Radios & backhaul | Smart agriculture, Meshtastic, and LoRaWAN, Field network IoT comparisons (HaLow, LoRa, NB-IoT), LoRa, MQTT, and gateway bridges; concepts LoRa (radio), LoRaWAN, Meshtastic, Wi-Fi HaLow |
Literacy and source clusters; not a chosen backhaul design |
| Edge compute | ESP32, ESPHome, and smart-farming builds, Homelab, self-hosting, and edge narratives, Docker, Kubernetes, Compose, and Bake (edge and homelab) |
Patterns; no farm edge boundary (what runs in field vs barn server vs home) |
| Home Assistant / farmOS / MQTT / TSDB roles | Farm data, farmOS, and agriculture lab builds (farmOS + lab); Smart home — Matter, Thread, and Home Assistant AI; Data storage (farm and edge stacks) (MQTT + PostgreSQL/PostGIS documentation links); PostgreSQL/PostGIS source-notes |
Roles not adjudicated: no system-of-record decision; TSDB named as placeholder in telemetry draft only |
| Time sync | Time synchronization — NTP and PTP sources; concepts NTP, PTP; Timing on the farm — synthesis |
Good primer layer; not tied to sensor timestamp requirements per subsystem |
| Device identity | Implied in Spatial data and farm asset registry standard (telemetry device_id); MQTT topic patterns in LoRa/MQTT source-notes |
No global naming authority document; no MQTT topic tree owned by you |
| Asset registry | Spatial data and farm asset registry standard (draft); PostGIS / enterprise PostGIS source-notes under Farm data… |
Draft standard; not populated with real IDs |
| Alerting | Automation degraded modes and manual fallback SOP (failure classes); HA/MQTT threads in topics |
No alert routing design (who gets paged, escalation); no “alert pipeline” SOP |
| Observability | Backlog P2 #19 (Observability and alerts — broker down…) — not a standalone page | Gap |
| Firmware lifecycle | Backlog P2 #20; ESP32/Meshtastic culture in topics | No policy for update cadence, staging, rollback |
| Degraded modes | Automation degraded modes and manual fallback SOP (first draft) |
Needs system-specific fill-ins and tabletop validation |
| Cybersecurity | Backlog P2 #15–16 (OT security concept + farm cybersecurity hub) — hubs not created | Gap; scattered homelab only |
| Remote access | Homelab topic; no bastion/VPN architecture for field admin | Gap |
2. What is missing (architecture-critical)¶
Aligned to audit §3–5 and backlog P0–P2 for technology.
| Layer | Missing artifact | Why it matters |
|---|---|---|
| End-to-end reference | One maintained reference architecture (diagram + dataflow + SPOF) approved for your sites | Without it, every purchase is ad hoc |
| System of record | Analysis: Telemetry system of record — options and boundaries (farmOS vs HA state vs TSDB/DB) — backlog P0 #6 | Conflicting truth breaks automation and records |
| Integration pattern | Canonical event → device → asset → map geometry path | Audit: farmOS and telemetry are adjacent, not integrated |
| Topic hubs | Field telemetry and backhaul, Farm cybersecurity and remote access (backlog #5 topic + #16 hub) |
Stops knowledge living only in long analyses |
| Security | OT/security concept + remote access story (VPN, segmentation, MQTT TLS/auth, camera exposure) | Remote farm is a high-value target for misconfiguration |
| Operations | Observability page; firmware lifecycle page; broker-down / false-positive runbook | MTTR and trust in alerts |
| Time | Per-subsystem clock requirements (e.g. sub-second unnecessary for soil moisture; correlation for triangulation)—not written | NTP/PTP exist but not bound to architecture |
| Comparisons | Gateway architecture — fixed tower vs vehicle vs mesh; farmOS vs lightweight (partially economic, partially ops) | Decision-grade tradeoffs |
3. Where the architecture is currently fragmented¶
Cross-cutting breaks (from audit §4 Missing integration layers, narrowed to technology):
- Radios ↔ land: LPWAN/mesh topics do not consistently reference paddock IDs, gates, or terrain—
Domain content overviewstates this gap. - Records ↔ telemetry:
Farm data, farmOS…and MQTT/LoRa topics do not define a single ingest or ID scheme linking observations to farmOS assets. - Homelab ↔ field:
Homelab…andSmart home…are not tied to farm alerting, on-call, or degraded modes at the 120-acre site. - Power ↔ compute ↔ radio:
Off-grid solar power and storageexists; field gateway uptime vs battery days is not one diagram with telemetry. - Time ↔ events: NTP/PTP strand does not state which logs must be traceable for debugging vs legal vs nice-to-have.
- Security ↔ remote access ↔ actuators: Irrigation/gates appear in degraded-mode SOP placeholders; no threat model (who can open what remotely).
Symptom: Strong horizontal coverage (many topics); weak vertical slice from sensor → identity → storage → alert → human → fallback.
4. Reference architecture page outline (target: one canonical page)¶
Suggested path: wiki/analyses/architecture-reference-smart-farm-homestead-and-120ac.md (or evolve Field telemetry reference architecture… into this v1.0).
| Section | Contents |
|---|---|
| 1. Document control | Version, owner, last site walk date, next review (quarterly recommended) |
| 2. Sites & trust boundaries | SITE_HOMESTEAD vs SITE_FARM; LAN, VPN, cellular, internet; what must never cross (e.g. cameras to public MQTT) |
| 3. Logical diagram | Field devices → edge/gateway → backhaul → broker → consumers (HA, farmOS ingest, TSDB, alerts) |
| 4. Dataflow & system of record | Authoritative store per data type (telemetry history, agronomic record, alert state); conflict resolution |
| 5. Identity | Device ID, MQTT topic root, TLS/client ID policy; link Spatial data and farm asset registry standard |
| 6. Time | Stratum for servers; NTP sources; PTP if used; sensor timestamp rules |
| 7. Radios & backhaul | Chosen stacks (placeholders until decided): LoRaWAN vs Meshtastic vs HaLow vs Wi‑Fi; failover |
| 8. Edge compute | What runs on gateways vs barn NUC vs home; container runtime if any |
| 9. Alerting & observability | Path from metric/event → notification; on-call; dead-man checks for broker |
| 10. Degraded modes | Link Automation degraded modes…; per-system safe states |
| 11. Security & remote access | VPN, MQTT auth, segmentation, camera policy, patch cadence |
| 12. Firmware & lifecycle | Update policy, test device, rollback |
| 13. SPOF & test plan | Table: component, failure mode, detection, mitigation, last drill |
| 14. Related diagrams | List of required figures (below) |
5. Required SOPs (prioritized)¶
| SOP | Intent | Backlog / status |
|---|---|---|
| Degraded modes — automation off | Manual safe state when logic or comms fails | Draft: Automation degraded modes… |
| Observability — broker down, false positives, on-call | What to do when alerts lie or stop | P2 #19 — to write |
| Incident — telemetry integrity suspected (sensor lying, stale data) | Disable closed-loop control | New |
| Firmware / certificate update — field gateway | Ordered steps; rollback | Derive from P2 #20 |
| Remote access session — admin to field network | VPN/bastion steps; no permanent port forwards without review | Part of cybersecurity hub |
| Time sync failure — skew beyond threshold | Which subsystems halt vs degrade | New |
| New device onboarding | ID assignment, topic provisioning, registry row, test plan | New |
6. Required diagrams (checklist)¶
| Diagram | Shows |
|---|---|
| D1 — Physical topology | Homestead vs farm buildings, towers, gateways, solar, fiber/radio paths (even notional) |
| D2 — Network & trust zones | VLANs or zones: cameras, PLC/irrigation, guest Wi‑Fi, management |
| D3 — MQTT / dataflow | Topics (high level), broker, bridges, HA, farmOS, TSDB |
| D4 — Alert path | Trigger → rule engine → channel (SMS/push/voice) → escalation |
| D5 — Power & uptime | Field gateway: solar/battery/grid, autonomy days, low-voltage cutoff |
| D6 — Time | Which hosts use which NTP/PTP; GNSS if any |
| D7 — Degraded mode decision tree | Internet down / broker down / sensor fault / actuator fault → safe state |
| D8 — Remote admin | VPN to where; jump host; no direct SSH to cameras from internet |
Format: Mermaid in wiki, or exported PNG/SVG in raw/assets/ with source-note—your choice; keep one canonical diagram set per AGENTS.md citation rules.
7. Related pages¶
Implementation backlog — strategic audit (P0–P3)Information architecture — decision-safe operational brain (target design)Business viability and farm economics — gap analysis— economics only
Update when P0 telemetry architecture is promoted from draft to v1.0, or when cybersecurity / observability hubs ship.