ComfoClime unit becomes unresponsive after 1-2 days

[helgeklein]

Since I've enabled this integration, my ComfoClime unit becomes unresponsive after a while. There doesn't seem to be a pattern, and the length of time after which it happens varies, but typically after 1-2 days.

When this happens, I the ComfoClime doesn't even repsond to pings anymore. The ComfoClime app also shows the unit as disconnected. I then press the reset button on the ComfoClime's Arduino board.

Do you have an idea what might be causing this?

[tokylo]

I've noticed the same thing: after 1-2 days, my ComfoClime unit also stops responding. I unplug the 230V power supply, and then it works again.

[helgeklein]

Thanks for confirming, @tokylo. I'd be happy to run tests, but I'd need some direction from @msfuture what to try.

[msfuture]

Same here, the unit becomes unresponsive after about 1-2 days.

That's a kind of difficult problem, as there isn't any error message (at least I didn't see any). For the meantime you could create an automation to restart the unit for example at midnight:

alias: restart ComfoClime
description: ""
triggers:
  - trigger: time
    at: "00:10:00"
conditions: []
actions:
  - action: comfoclime.reset_system
    data: {}
mode: single

This works in the latest beta, where the restart action has been added.

[msfuture]

Possible reasons could be:

• polling connections are too frequent and thus the HTTP interface locks up (we can check if the CC keeps communicating with the ComfoNet Bus)
• polling connections aren't terminated correctly (we could listen on the communication and checking for TCP FIN)

There is a chance, that this is a firmware issue of Zehnder. If that's the case, I'm not very positive that they will fix it. I reached out to the support and they are denying any communication with the developer team, as there won't be any support of the API.

I currently try to throttle the polling by delaying each request by a second. There already is a locking structure in my code, that prevents parallel requests to the unit, but maybe it needs some time between the requests to handle other threads.

[helgeklein]

Thanks, @msfuture. I'll be happy to try the restart action once my PR has been integrated into the beta 😉

I agree that it's likely that the frequency of the API calls is too high.

I also reached out to Zehnder and got basically the same reply: the ComfoClime API is unsupported. I think it's important, though, to let them know there's interest. Maybe they change their mind at some point.

[msfuture]

Maybe I find some free time on the weekend to integrate the PRs.

I'm not 100% sure about this; Maybe the Python script is not always closing the connections. A call to response.json() should close the connection, but if an error happens it may be left open. I will do further investigations on this.

[msfuture]

The workaround with the reboot at midnight is working for me in the meantime. I'm currently monitoring the TCP connections for a few hours to see if they are terminated properly.

[helgeklein]

I've just created the daily reset automation of the ComfoClime device. I'll let you know if that works around the issue.

[msfuture]

I had some issues today again. But after about 30 minutes the CC unit came back online. Doesn't look like it totally locked up.
The dashboard / thermalprofile queries cannot be the problem. But all property and telemetry queries are sent out really fast. Currently I prepare a PR that delays queries for about 300ms before sending out the next one and ensures the HTTP connection is terminated properly. We have about 30 single queries, 300ms should not be a problem then and HA handles this async anyway.

For the properties there seems to be a possibility to query multiple props in one query. But alle my attempts with postman fail. It should be the /method/X/Y/2 endpoint (2 is for querying multiple properties). I tried to post data like in https://github.com/michaelarnauts/aiocomfoconnect/blob/master/docs/PROTOCOL-RMI.md but hat no success yet.

For the telemetries I didn't find any possibility to retreive multiple in one query.

[Puekcommander]

The connection issue was resolved in firmware version 1.4, as it had been quite unstable in earlier versions. Unfortunately, the problem has reappeared in version 1.5. I’ve contacted Zehnder to address both this issue and the malfunctioning free cooling mode. As you're likely aware, the ComfoClime will not cool when the bypass is active (disabling the bypass allows cooling via the ComfoClime).

Is there any way we can download the firmware ourselves to inspect what's been changed or included?

[msfuture]

The nightly restart seems to improve the issue most of the time. I think that maybe it isnt the API that freezes but the ESP losing the WiFi Connection...

[helgeklein]

Since I enabled the nightly restart the issue hasn't happend anymore.

[msfuture]

We could think about triggering the reset automatically in the integration when the uptime hits a certain value. But as this is only fixing the symptoms and doesn't care about the reason, I would like to do some further testing in the codebase.

[helgeklein]

I totally agree. Fixing the root cause would be very much preferable.