Huawei SUN2000-6KTL-M1 SDongle Modbus TCP: "Reading from inverter failed" / register 30000

Configuration
,
1

I’m trying to connect a Huawei SUN2000 to Home Assistant locally over Modbus TCP via SDongle, but setup always fails
at the initial inverter read.

Setup:

  • Inverter: Huawei SUN2000-6KTL-M
  • Communication device: SDongleA-HV2150154700
  • Dongle IP: 192.168.1.41
  • Modbus TCP is enabled on the Huawei side
  • Connection type: Network
  • Main tested port: 502
  • Fallback tested port: 6607
  • Tested slave IDs: 1 and 0
  • Home Assistant version: 2026.2.2
  • Integration tested:
    • Huawei Solar 1.6.0
    • beta 2.0.0b2

What is confirmed:

  • 192.168.1.41 is reachable
  • 502/tcp is open
  • 6607/tcp was previously closed/refused
  • HACS / Home Assistant side is working correctly
  • The integration loads correctly and appears as a valid config flow handler
  • No duplicate custom component copies remain

What happens:

  • The integration never creates a config entry
  • Setup fails during the very first device identification read
  • On stable I got read_error
  • On beta 2.0.0b2 I now get unknown, but logs are more specific

Relevant logs from beta:

  • Connection failed when trying to read registers model_name
  • Connection error while reading registers [‘model_name’]
  • tmodbus.exceptions.ModbusConnectionError: Transport is not connected
  • Received an ModbusConnectionError error. Closing the connection to force a reconnect.

Earlier stable logs also showed:

  • Failed to read register 30000 after 6 tries

  • Modbus client is not connected to the inverter

  • 192.168.1.41:502, slave 1

  • 192.168.1.41:502, slave 0

  • beta and stable

  • all with elevate permissions = off

All of them fail before the inverter model can be read.

At this point it looks less like a Home Assistant installation issue and more like a Modbus TCP / SDongle / firmware /
active-session problem on the Huawei side.

Question:

Has anyone seen this exact behavior with SUN2000 + SDongle, where TCP is reachable but the connection drops already on
model_name / register 30000 reads? If yes, was the root cause:

  • another active Modbus client / Device Commissioning session
  • SDongle firmware
  • inverter firmware
  • port / slave-id specifics
  • or something else?