Intermittent Connection failures due to Chamberlain MyQ API transition

[GaryOkie]

The latest connection failures are no longer related to V5.1 user agent changes but are evidently due to Chamberlain transitioning to a new V6 infrastructure and API. Please refer to home-assistant/core#44851 for further description of the connection errors.

Work on reverse engineering V6 has already been started by the Homebridge-myq developer (which does not use pymyq). He is encouraging discussion on Homebridge Discord so please consider a collaboration on this effort.

hjdhjd/homebridge-myq#226 ...

we’re aware there’s been yet another change in the myQ API and infrastructure. They seem to be transitioning to a new infrastructure and the errors we’re seeing are a result of the transition process and that they are deprecating the prior APIs we’ve used. I’ve gotten some good logs of the new protocol...now it’s a function of implementing it over the coming days. Stay tuned.

[mikepayne02]

Sounds about right. My instance still spits out a 'Server Error' intermittently even with the user agent reverted back to 'pymyq.'

[brbeaird]

This is actually a relief - my SmartThings integration has also been seeing intermittent 500 errors for the last several days. Was waiting to see if an issue would pop up here. Very glad to see someone has already started on V6.

[haffi78]

Any chanse that someone might look to see if we can support the now one year later the lockitron locks wich show up on the app also ? Its just show as another type of door?

[wpnelson]

Thanks to any and all who contributed to the quick fix.

[FreeWillyB]

To whoever can help... I’m trying to reverse engineering this new API, but I’m stuck at the starting blocks. It appears you can only control the devices from an app, so I can’t use chromes inspection tools. I’ve tried a MITM attack, and after ALL NIGHT, I was finally able to get it somewhat working with mitmproxy and adding a trusted CA on my iPhone. But even though that worked on somethings, others like MyQ still didn’t work properly. Any ideas on how I can dig into this and grab some of the info on this new API?

[SweVictor]

Thought I'd help but got to the same place as you @FreeWillyB. Basically most apps today seem to use certificate pinning, and to capture that traffic you need a certificate (MITM cert for example) in the SYSTEM store, not USER store (user store is disregarded for apps since v7 of Android but used by apps explicitly opting in, chrome for example).

App traffic can probably be captured in an ADB environment or (for sure) on a rooted device, but I have neither up and running so can't help unfortunately.

Edit: Aaaand after re-reading first post work on gathering API calls has probably already been done in the homebridge repo....

[mikepayne02]

Today around 9:15 AM CST the intermittent errors and unavailability un Home Assistant stopped happening. MyQ may have fixed a bug or changed something, but it seems to be working okay for now. V6 api definitely needs to come asap because who knows how long this will keep working.

[GaryOkie]

Are you from the future? It's 8:45AM CST here in Central US now :)

Anyway, I concur that these errors have stopped for me as well - but as of yesterday @ 9:08PM CST.

[garyak]

My last disconnect was at 11:54 pm MST.

[mikepayne02]

Not a time traveler my bad, I meant 9:15 PM CST, which matches up with what you noticed as well.

[juched78]

Same here, disconnects stopped for me around 9:45PM EST yesterday.

[ds7771]

Same here, my logs show disconnects stopped at 9:57PM EST Jan 12.

If you are following the Homebridge project, they just released a patch to repair v5 API as a temporary measure. But it is unclear to me if the success of their patch is really due to MyQ API working again as of last night.

Anyway, just pointing out, if you want to run HB in parallel with HA, then you can bring the working Homekit devices from HB into HA and use those entities in your automations (assuming their patch is a real fix).

[brbeaird]

But it is unclear to me if the success of their patch is really due to MyQ API working again as of last night.

I have reviewed their commit and from what I can tell, it does not actually change the way it calls the API. I do believe changes on the MyQ side are the real difference-maker.