SG15.RT: no "work_state_1", kills exporting to influxdb

[pohly]

I recently had an incident where a SunGrow SG15.RT entered "state: not initialized" (from its web interface) and no longer produced anything. Power cycling it fixed this. I now checked whether I can store that state in influxdb to set up some alerting solution.

I found "work_state_1" and "work_state_2" in the registers file and added that:

  # Output data to InfluxDB
  - name: influxdb
    enabled: True                           # [Optional] Default is False
    ...
    measurements:                           # [Required] Registers to publish to bucket
      - point: "work_state_1"
        register: "work_state_1"
      - point: "work_state_2"
        register: "work_state_2"
      - point: "energy"
        register: daily_power_yields

This didn't work. Debug output shows that only "work_state_1" is present, but with None as value:

Jan 07 14:42:10 xbmc conmon[496320]: +----------------------------------------------------------------------+
Jan 07 14:42:10 xbmc conmon[496320]: | Address | Register                            | Value                |
Jan 07 14:42:10 xbmc conmon[496320]: +---------+-------------------------------------+----------------------+
Jan 07 14:42:10 xbmc conmon[496320]: | ----    | device_type_code                    | SG15RT               |
Jan 07 14:42:10 xbmc conmon[496320]: | vr001   | run_state                           | ON                   |
Jan 07 14:42:10 xbmc conmon[496320]: | vr003   | last_reset                          | 2024-1-7 14:42:09    |
...
Jan 07 14:42:10 xbmc conmon[496320]: | 5038    | work_state_1                        | None                 |

Then exporting to influxdb dies:

Jan 07 14:42:10 xbmc conmon[496320]: Traceback (most recent call last):
Jan 07 14:42:10 xbmc conmon[496320]:   File "/usr/src/sungather/./sungather.py", line 190, in <module>
Jan 07 14:42:10 xbmc conmon[496320]:     main()
Jan 07 14:42:10 xbmc conmon[496320]:   File "/usr/src/sungather/./sungather.py", line 158, in main
Jan 07 14:42:10 xbmc conmon[496320]:     export.publish(inverter)
Jan 07 14:42:10 xbmc conmon[496320]:   File "/usr/src/sungather/exports/influxdb.py", line 64, in publish
Jan 07 14:42:10 xbmc conmon[496320]:     value = inverter.getRegisterValue(register) if type(inverter.getRegisterValue(register)) is str else float(inverter.getRegisterValue(register))
Jan 07 14:42:10 xbmc conmon[496320]:                                                                                                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Jan 07 14:42:10 xbmc conmon[496320]: TypeError: float() argument must be a string or a real number, not 'NoneType'
Jan 07 14:42:11 xbmc podman[496331]: 2024-01-07 14:42:11.01818019 +0100 CET m=+0.046560191 container died e41c7fefdc2a2af0a8bd1dae72c13f7e80cef0ce02a083e94b77a3913ade42e1 (image=docker.io/bohdans/sungather:latest, name=sungrow-15)

I suppose the influxdb code should handle None more gracefully.

Not sure whether I really want work_state_1, though. Perhaps run_state is what I am looking for.

[pohly]

Looks like I need work_state_1 after all. The inverter is down again, with:

vr001   run_state   ON 
5038    work_state_1    None 
5081    work_state_2    None 

[pohly]

I power cycled the inverter. Now I get something other than None:

vr001   run_state   ON 
5038    work_state_1    Run 
5081    work_state_2    Stop 

So it looks like it's only the "uninitialized" work state that isn't handled. Fixing the handling of None in influxdb.py:64 such that missing values get skipped should be enough for me: then I can set up alerting with "unknown state for xxx minutes" as trigger.