TCP Client & Server - S2-WL-ST Troubleshooting and Config changes

[solarjones]

I've been running 2x Waveshares together with a per the design for just over 2 months and during that time have had a few issues.

The Solis data stick I've used since installation (S2-WL-ST) with Solis Cloud had been flawless during that time until very recently where it went offline. I power cycled my PoE hub which brought it back online briefly, but in the end had to unscrew the data stick to power it off entirely (which makes sense since it would still receive power directly from the inverter). This is the first time the Solis data stick has 'hung' since installation.

Previously the anomalies have been with the HA data spiking on battery SOC, export & import values, which in turn produces bad energy stats.

I'm running Solax Modbus 2023.05.2b3 with the following configs on Waveshare:

Server

Client

The pattern I've observed is everything works well for a few weeks then hits an issue.
Any suggestions?
Thanks.

[alienatedsec]

@solarjones

I would call datalogger disconnections a norm even without our unique 2x Waveshare setup. When you read the FB group, you can see many people doing the same thing - getting to their lofts and reconnecting (restarting) their dataloggers. However, there are several things you could introduce as an improvement.

I recently started using the same datalogger - see #7 (reply in thread)

I had no problems with the Solis Cloud and the datalogger since I swapped my dataloggers from DLS-W to S2-WL-ST - its firmware is 10010115

You could also tick Restart for no data within the TCP Client config - if you had to power cycle the datalogger, it could have frozen as it is known with other cases, and looking at only one instance is not worth any further troubleshooting.

Previously the anomalies have been with the HA data spiking on battery SOC, export & import values, which in turn produces bad energy stats.

To mitigate the HA data spiking, I have taken another approach and amended my inverter and TCP Server serial settings to a higher baud rate. See wills106/homeassistant-solax-modbus#340 (comment)

I suspect there was a bottleneck with the default 9600 setting on the inverter to support HA and the datalogger at the same time.

[alienatedsec]

It feels like a never-ending story with those spikes and config combinations. Regardless, the outcome is always the same. Sooner or later spikes are happening and I am out of ideas as to which parameters to amend next. The bus conflict detection and its idle help, but is not enough. I guess, I will keep increasing it until I find the right setting or analyse the traffic. In the end, it could be how pymodbus is handling data but too early to confirm.

[solarjones]

Have you had the spike (SOC 0%) in the other integration too and was it in parallel with Solax integration?

Also wondering if the integration needs auto reloading once a week, for example to clear any variables it's using internally that perhaps overflow causing a crash.

[alienatedsec]

@solarjones it was only a short-term test. I had to disable the other integration as it handles queries differently from SolaX. SolaX is reading blocks of data which is quick, Francis’ method is to get individual readings. That was slowing down the overall performance of both integrations.

I had a spike this morning but that could be me increasing the block size to 80 rather than leaving it at 40. Experimenting with too many things is not good sometimes.

For now, I left all as above and loaded the latest SolaX version. That is back to 40 blocks of testing. I'll keep you posted.

[wills106]

I had a spike this morning but that could be me increasing the block size to 80 rather than leaving it at 40. Experimenting with too many things is not good sometimes.

The Modbus Docs for Solis states the max amount of registers to read is 50 16bit registers. So Block size needs to be kept under 50.
Running a blocksize of 48 would reduce the amount of requests to the Inverter.
Running a blocksize of 30 would increase the traffic to the Inverter. But it would only have to respond with a smaller dataset each time.

I don't know if there is a happy medium?

[alienatedsec]

@wills106 It has been another day of me messing around. I am unsure if this is a Waveshare issue, but regardless of the settings and the amendments I made, the integration failed and spikes were regular.

It looks like the only stable setting is the default setting even if I enter the same values as default values, it won't work - Now, I can see why it keeps failing no matter of amendments and values that we believe should help in our case.

I raised a ticket with Waveshare as their device should support up to 30 connections, but its performance is questionable with 2 and unacceptable with 3 connections - I know that Solis could have a big part in that poor performance. Let's see what they come back with, but it could be several days before I get the answer as was the case in my previous ticket.

[alienatedsec]

@wills106 FYI

@solarjones Waveshare came back to me and shared a new firmware version v1.486, and that is a few releases ahead when compared to my old one v1.452

Flashing the firmware also changes the MAC address, so if anyone wants to try it - do it at your own risk. There are two parts - the firmware part, which is wiping the web interface, and the web interface to be flashed after.

Here is the link: https://note.youdao.com/s/JImk2MbZ

[EDIT] Following the successful testing, I also created the prerequisite guideline here #17

[solarjones]

Thanks for sharing. Did Waveshare provide any release notes or indicate an issue that's been addressed in the latest firmware?

[alienatedsec]

@solarjones Waveshare came back with 4 points to advise on the problem. I had to rule out two of them and one is gone back with further questioning as they mentioned a parameter config field to amend that does not exist. Once, I have more clarity, I will share the outcome. The last one was the firmware update.

[alienatedsec]

I initially found that the new firmware is working better, and has not failed for a few hours on multi-host non-storage type settings. The auto-query storage type keeps failing regularly. It would be against the manual, but I am keen to leave it that way unless it will suddenly fail.

[alienatedsec]

@solarjones No issues since yesterday's change to multi-host non-storage type on TCP Server with the new firmware. I'll give it a week before updating the guideline.

Did you change anything on yours?