Essentially, I’ve taken the core integration and modified the way in which product information is initially enumerated for each bulb as well as the way in bulb updates are done. I’m currently running this for my fleet of 60 bulbs and have dropped from hundreds of timeouts to zero, so now I need to find out if this is reproducible for other folks.
However, I don’t consider this a stable release, and would only suggest this to folks who are comfortable adding/removing custom components that override the default integration. It should be a drop-in replacement and removal would revert you back to stock, but if you’re not comfortable with potentially having your lights not work as expected, I don’t mind if you skip it.
I have a couple of LIFX bulbs that are in non-critical locations, so I can try. However, I’ve not overridden a default integration before. Do I just install from HACS (comfortable with this), or to I have to also delete the integration? If the latter, do I also have to then add the integration back using the one that HACS would define?
You just install using HACS after adding the GitHub repo manually. HACS won’t add custom components with the same name as core components to the default list, which makes sense. Once installed, restart Home Assistant (as with any HACS integration install) and It’ll automatically override the core integration and use the existing configuration. It should “just work”.
If it doesn’t “just work” for you though, uninstall with HACS and restart Home Assistant and it’ll go back to using the core integration. You can tell the difference by checking the logs: the core integration will log using homeassistant.components.lifx while the beta uses custom_components.lifx.
Yeah, I didn’t want to make this an auto-updating release in case someone was still running the old version and upgraded without warning. This really had to be super opt-in.
btw, I’m interested in feedback both objective (i.e. less logged errors or warnings) and subjective (you feel your lights react faster/slower or are more or less online than before).
A bit hard to see, but the upshot is before 7pm rubbish, after 7pm rock solid. Guess when I installed the beta? I normally only turn these on and off with automations so response is not an issue, but standing beneath with the companion app they respond immediately - faster than my Hues. Well done!
Right now? If you’re running 2022.11, they’re about as stable as the core integration, maybe slightly higher. And it’s fairly easy to back out, even remotely: just delete/uninstall the custom component.
However, if you’re planning on upgrading to 2022.12 or are running the dev branch already, there are some fixes in dev for the update coordinator that are the real fix for some of the things I was hamfistedly trying to workaround, so now I’m working on re-factoring my changes on top of those fixes. That will take some time, because in the process I appear to have broken almost all the tests which is both impressive and tedious.
Yeah, @bdraco’s discovery and fix for a thundering herd problem has made a noticeable difference in flattening out the connectivity curve for my fleet of 60 devices. Now I’m trying to add a bit more resilience for when a bulb just doesn’t respond for over 8.5 seconds.
I’m also playing with some in-flight limits so as not to overwhelm devices with compounding requests.
I’m extremely pleased with how my refactor for 2022.12 is performing. It has dropped integration start time for my fleet of 60 bulbs down to 1-2 seconds and I’ve also been able to increase sensor updates (like RSSI) threefold from every 30 seconds to every 10 seconds. All that with no devices going unavailable at all (though testing has been limited).
I’ll merge the refactor and release it as LIFX Beta 2022.12.0.dev0 when Home Assistant 2022.12.0 is available in the beta channel. The refactor is so significant that almost none of the current tests are working. I’ll be working on updating the tests for the refactor while hopefully getting real world feedback from my loyal army of crash test dummies courageous volunteers.
A whole new year and a whole new beta: 2023.1.0b1 is now available and reverts most of the integration back to the Core version and uses a new method to improve stability I hope is as effective for everyone else as it appears to be for my environment.
I’m not on Discord very often. If you have any issues with the LIFX Beta, it would be great if you could open an issue on GitHub for it: Issues · Djelibeybi/ha-lifx-beta · GitHub
Been on 2122.12.6 for a little while and had just a couple of issues with one bulb. Will see how this one goes. Still hands down better than the standard integration.
I’d be keen to test. I’m new to HA, and setting up a new home assistant instance. I have a few Lifx bulbs as well as a couple of switches. Can’t get homekit working on them, so i’m also keen to understand if you’re planning to integrate photons natively so we can get lan control of switches without homekit or cloud ?
Photons doesn’t provide LAN control for Switches, so that’s not an option. The only way to get LAN control is via HomeKit, either natively or using the HomeKist Controller integration.
(Well, mostly: you can control the relays/terminals locally via Photons, but there is no visibility to the buttons, which means nothing is notified if anyone pushes any of them, except either the LIFX Cloud or HomeKit).
I’ve been running it on my main instance for the last couple of days and haven’t had a single bulb go offline once. In fact, the only log entry I have for LIFX is Home Assistant complaining once that a single bulb took 0.844 seconds to respond and that I should report it. So I’m reporting that improvement to all of you.
Considering it’ll be a while before that code makes it into a stable release, installing the beta via HACS is a great way to test it now before it does, without having to opt-into Home Assistant’s beta stream.
Note that the Beta does generate warning log entries that are not actual warnings: they are debug logs in the core. I just made them warnings to make them more obvious when scanning the log files. Not that I’ve seen any recently anyway, but if you do, please let me know!
