Latest HA 2023.9.0 breaks SolArk Integration

[poldim]

Everything worked fine until I updated to 2023.9.0, reverting to 2023.8.4 allows it to work again.

This error originated from a custom integration.

Logger: custom_components.solark_modbus.hub
Source: custom_components/solark_modbus/hub.py:114
Integration: SolArk Inverter Modbus (documentation, issues)
First occurred: 8:28:47 PM (51 occurrences)
Last logged: 8:32:57 PM

Reading realtime data failed! Inverter is unreachable.

I've scraped the changelog for anything modbus related, two stand out:

• Modbus switch, allow restore “unknown” (@janiversen - #99533) (modbus docs) (beta fix)
• use write_registers also for target temp (@brunoenten - #97475) (modbus docs)
• Bump pymodbus v3.4.1. (@janiversen - #97612) (modbus docs)
• modbus: use pb not pymodbus consistently as name. (@janiversen - #97780) (modbus docs)
• Fix Flexit mypy error in pymodbus (@janiversen - #97799) (flexit docs)
• Handle explicit Modbus NaN values (@joanwa - #90800) (modbus docs)
• Replace Float ‘nan’ with None for modbus floats (@String-656 - #93832) (modbus docs)
• modbus: remove unused constants and get 100% coverage. (@janiversen - #97779) (modbus docs)
• Move all used modbus constants to Stiebel (@joostlek - #98044) (stiebel_eltron docs)
• Make changes in modbus trigger a full CI run (@emontnemery - #98055)
• Modbus: set state_class etc in slaves. (@janiversen - #98332) (modbus docs)
• Address late modbus review (@janiversen - #99123) (modbus docs)
• Add possibility to have multiple values for every modbus hvac mode (@escoand - #98829) (modbus docs)
• Add modbus test for configuration errors (@janiversen - #98697) (modbus docs)
• Modbus switch, allow restore “unknown” (@janiversen - #99533) (modbus docs) (beta fix)

[janiversen]

You need to ask the custom component to update.

[legomind]

can confirm this problem exists in 2023.9.0. Redownloading from HACS does NOT fix the issue.

[pbix]

You need to ask the custom component to update.

What update is required? I do not see what exactly has broken.

I saw all these modbus related changes in the changelog for HA and decided to wait a bit before upgrading my system even to 2023.8.4 (many modbus related regressions had to be addressed in the changelog I saw due to changes in 8.0 it appeared to me).

Anyway I have not yet had a chance to investigate and would welcome any input.

[legomind]

@pbix is there any way to get the details of the exception logged?

The exception being raised is caught here: https://github.com/pbix/HA-solark-PV/blob/9df26260e0f3f14939fff97151d359bc6b952b3a/custom_components/solark/hub.py#L113C10-L113C10

But the underlying error is not actually logged, as far as I can see.

[pbix]

Possible regression in 2023.9.0. Please avoid updating for now and let's see what becomes of this upstream.

home-assistant/core#99784

[pbix]

Looks like regression #1758 in pymodbus 3.5.0 is responsible for this.

Home Assistant 2023.9.1 is slated to move to v3.5.1 to address this.
home-assistant/core#99940

I think there will be a new HA release soon so please hang on a little longer.

A confirmation by anyone upgrading to 2029.9.1 or later in this thread would be appreciated so I can close this issue.

[dscowan]

I upgraded to 2023.9.1 and it is still broken

[EiNSTeiN-]

Same, running 2023.9.1 and I'm still getting a Inverter is unreachable exception at solark_modbus/hub.py:114

[legomind]

Wait for 2023.9.2. It has the actual fix in it.

[alexdelprete]

Hi @pbix,

sorry to chime in, but I thought it might help: for my component the problem was byteorder=Endian.Big. Had to change all those to byteorder=Endian.BIG.

The change in pymodbus 3.5.0 was documented here: https://github.com/pymodbus-dev/pymodbus/blob/dev/API_changes.rst

Hope it helps.

[pbix]

Hi @alexdelprete,

Thanks for the tip.

Did you have to version detect to use lower case for versions prior to 3.5.0 and upper case for later versions?

What a pain.

[alexdelprete]

Hi Paul,

Did you have to version detect to use lower case for versions prior to 3.5.0 and upper case for later versions?

No, I decided to force the use of HA 2023.9.1 and pymodbus 3.5.1 so I could immediately release a working version. I hope to not cause too many problems to users. I'll wait for feedbacks and then, if needed, I might implement what you suggested. I saw your code, and I noticed you already check pymodbus version for other things, it's a good idea.

What a pain.

Absolutely, I spent many hours debugging this (mainly because I'm not really a developer) and I found the solution only because I clicked on the right link and when debugging, examining the exception I was receiving, it contained the string "Big". So I associated Big to that Endian case change, and that was definitely it. A little bit of luck and years of experience in troubleshooting IT issues. But you're right: "what a pain..." :)

[pbix]

Update to Main branch allows this integration to operate despite the breaking changes introduced without warning into the pymodbus library. The SolArk integration should now work with a releases of HA.

Thanks to @alexdelprete for pointing out one of the breaking changes saving me many hours if debug no doubt. Your integration may benefit from my resolution method as it works with all versions of HA. Just avoid using the pymodbus Endian module altogether.

Would love to hear from other users that they met success after updating to this release.

[EiNSTeiN-]

I've just updated (redownloaded from HACS to be sure) and the integration still seems broken after restarting HA

[alexdelprete]

Thanks to @alexdelprete for pointing out one of the breaking changes saving me many hours if debug no doubt.

glad to help. I hope that's the only issue we're facing: I have 3 users reporting it worked, but now 1 user tells me it's still not working. :(