bug: /status reports not populating stats for some clients

[t413]

Again. Incredible project. Love it!

I've added an observer and it's working perfectly, forwarding packets and populating under /#/observers/* nicely.

But the node stats aren't filled in from my node. The fields for Battery, Uptime, Model, Firmware, Noise Floor, etc are all just "--". My node, in particular is this one.

Other nodes, for example running RemoteTerm, populate this data perfectly. It shows the battery voltage and all other stats, and tracks voltage over time. But for my repeater / observer running EastMesh firmware those details are empty. The /status data posted to mqtt should be the same, but here's an example directly from my node, posted to meshcore/OAK/2DD2D5226F59371F092B203C93A9B76D23D32B1B4122896A89C4EEBF66B0821F/status

{
  "status": "online",
  "timestamp": "2026-06-17T08:42:53.000000",
  "origin": "t4 oak hills",
  "origin_id": "2DD2D5226F59371F092B203C93A9B76D23D32B1B4122896A89C4EEBF66B0821F",
  "model": "Heltec_V4_OLED",
  "firmware_version": "v1.16.0",
  "radio": "910.525024,62.5,7,5",
  "client_version": "EastMesh/v2026.6.3-235e540",
  "repeat": "on",
  "stats": {
    "battery_mv": 3458,
    "uptime_secs": 32457,
    "errors": 0,
    "queue_len": 0,
    "noise_floor": -76,
    "tx_air_secs": 16,
    "rx_air_secs": 1623,
    "recv_errors": 3714
  }
}

It doesn't seem possible for random users to subscribe to raw mqtt updates to compare the /status update of other nodes so I can't compare.. so it's up to you! Let's fix this!

[Kpa-clawbot]

Severity: P3 — observer metadata cosmetic, no functional impact.

Claim: confirmed (analyzer.00id.net DB shows this observer with packet_count=2454, last_seen=current, but model=null, firmware=null, battery_mv=null, uptime_secs=null, noise_floor=null; only client_version="EastMesh/v2026.6.3-235e540" and radio="910.525024,62.5,7,5" are populated).

Code today:

• cmd/ingestor/main.go:556-575 — status topic handler for meshcore/<region>/<observer_id>/status calls extractObserverMeta(msg) then UpsertObserverAt + InsertMetrics.
• cmd/ingestor/main.go:1046-1140 — extractObserverMeta reads top-level model, firmware_version, client_version, radio AND stats.{battery_mv,uptime_secs,noise_floor,...} (with top-level fallback). The reporter's payload shape (top-level model, firmware_version; nested stats.{battery_mv,...}) is exactly what the code accepts.
• cmd/ingestor/db.go:777-794 — stmtUpsertObserver uses COALESCE(?, model) etc., so a status update should overwrite NULLs.

Findings:

• (meshcore): payload shape matches extractObserverMeta exactly — top-level model="Heltec_V4_OLED" and nested stats.battery_mv would both be picked up. The decode path is not the bug.
• (carmack): observer has 2,454 packet observations from packet topic but zero successful status writes (every metadata column null). Either (a) the status messages are not being published to the broker we subscribe to, (b) they're going to a different topic shape than meshcore/<iata>/<id>/status, or (c) they're being filtered out earlier (IATA whitelist? observer blacklist?).

Fix path: needs operator confirmation before code change. Ask reporter to:

1. mosquitto_sub -h <their-broker> -t "meshcore/OAK/2DD2D5226F.../status" -v and confirm status JSON is actually published to the broker the analyzer is subscribed to (vs. a private/local broker only).
2. Confirm the topic shape — EastMesh firmware may publish meshcore/<region>/<id>/status or some variant (/stats?).
3. Check /api/observers for noise_floor populated on a known-good observer in the same IATA region (rules out region-wide subscription gap).

If status messages are confirmed reaching our broker on the standard topic and still not landing, then we need ingestor logs (MQTT [tag] observer status error: or absence of status logs) to identify the drop point. Effort: S (likely a topic/config issue, not code). Effort: M if a parser/case-sensitivity issue surfaces.

Dup check: searched "observer status battery null model" — no open dup; #1465 (status timestamp) and #336 (nested stats) are adjacent but different. PR #1682 added per-source status endpoints recently.

Triage verdict: needs-operator-input