docs(energy): note the meter converter's stray ACKs and Deye resets

Both show up in the Pi's journal and look like faults: packets from a public address on port 502, and Deye going down every 3 hours. Neither loses readings, so record them where the next reader looks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015x6FWeKCY59JfFJmzAm4TL
2026-10-25 14:30:53 +01:00
1 changed files with 8 additions and 0 deletions
@@ -93,6 +93,14 @@ House meter SDM630 (`electricity.PRAD2`): `total_power_active` is **+ import /
 Deye grid registers. Checked 2026-10-23 in InfluxDB: −2.5 to −5.5 kW between 11:00 and 15:00 local
 while PV exported, +645 W at night matching Deye's 640 W import.

+Link quirks seen from the gateway on the Pi (2026-10-24/25), both harmless to the readings:
+
+- The meter's RS485 converter (`192.168.0.10`, MAC `d8:b0:4c:bc:7a:8e`) sends late TCP ACKs from
+  port 502 with the source address `76.85.0.32` (TTL 255, so from the LAN, not the internet).
+  ufw drops them, about 1800 `[UFW BLOCK]` journal lines a day; the polls themselves succeed.
+- The Solarman logger resets our connection about every 3 h, always near :12
+  (`deye is down: Connection reset`); the next poll a minute later works again.
+
 Considered and rejected (2026-10-23): pointing the logger's second server ("Server B") at our
 backend. Push frames are poorly documented binary data every ~5 min, need a TCP server with
 acks, and a wrong setting could cut off the Solarman cloud my father uses. Local FC3 polling