These discussions about update philosophy are good to have. And the fact that they’ve been staying (mostly) polite is great.
There’s always going to be a tension between developers and users. Developers want to run on the latest and greatest hardware. They can imagine all kinds of cool features and want to add them all. They love to tinker and tweak, so when things break it’s a fun challenge, not something that ruins their day. They like speaking a secret language of jargon which excludes outsiders. Some even like obfuscating code to make themselves seem smarter.
Users just want things to work, quietly and in the background, without a lot of effort, without having to learn a bunch of jargon, without having to buy new hardware and without having to research how every “bump” change really impacts them.
Each group needs to understand - and listen to - the other if HA is to grow beyond a niche market of tinkerers.
TBH it became obvious to me once I got well down the rabbit hole of HA that I needed to make sure I had resilience in my system - via backups. If your system is mission critical then you shouldn’t be updating at all without a robust strategy in place. To me that means HA backups on device, in GoogleDrive and finally - importantly - image based backups (in my case via proxmox).
I can trial new releases and find problems and roll back - either to when I have time to deal with the problems or till they are solved in a later release. This way I’m also able to test for the project and raise issues or feedback for the team. I very much value the hard work on this project and I have realistic expectations. That said as others have pointed out we are all going to be using HACS for some stuff and that’s just something the project needs to live with. The ethos here is not to wall us in but allow us to play - and as end users we must accept some bumps on the road.
I’m a software engineer by profession for some decades (OK technically I’m management now) and the project(s) we work on are even more complex and very much hardware linked in terms of our product. This stuff ain’t easy and the team here are doing an incredible job.
Remember, you very likely don’t need to upgrade - at least not immediately. It’s handy if you do find problems though…
I like your point of view and thank you for your though, because I started thinking about what are the availability option today to achieve redundancy and high available. I’m happy to learn more if anyone would like to share it
I see your point. But yeah, the successful ones don’t require me to read hundreds of lines of release notes, then follow and understand links to highly technical Github threads. Or figure out whether or not a “bump” affects me. I generally just let my cell phone update without giving it a second thought. Many people do the same for Windows and the apps on it. I have multiple old machines and old apps which still work just fine with virtually no effort.
We who like fooling around with this stuff find it hard to believe, but most people will think of HA as sort of an appliance. Something you set up once and then it just works, doing what you need it to do, when you need it to. That should be the aspiration.
I have to call shenanigan’s on this. I do this for a living. We have to actively turn off all updates so that hardware continues to work because windows updates constantly break things. OS’s I manage: Windows NT, Windows XP, Windows 7, Windows 8, Windows 10. No windows 11 yet. Every single one has updates off because something breaks randomly when Microsoft pushes updates.
Recently (Last 12 months), Microsoft made speed improvements to COM communication in Windows 10 which flooded the buffer for some lasers that run on our assembly lines. The lasers simply couldn’t keep up with the buffer speed and there’s no fix for it from the laser vendor. We had to downgrade windows on all machines running the laser.
Dare I even bring up the whole Win10 FireWire (IEEE1394) deprecation?
EDIT: Oh, I had auto-updates on a Windows 10 box running BlueIris. The windows update borked the install and I had to wipe and reload. This happened in March, forgot about that. Needless to say, windows update service is disabled on that machine now.
If we’re being open to apples and oranges comparisons: my old car has required no updates ever and I didn’t need to read the manual to use it.
On a tangential note, people who recently allowed their iPhones to automatically update their Sonos app have learned they should have given it a second thought. (No rollbacks on iOS)
Typical open source software does not have the problems, Home Assistant has. At least not at this scale and frequency. Home Assistant is the only open source product I have a bad feeling with about every upgrade. You need to very carefully read the release notes and still need fear things breaking with every major update. Home Assistant feels more and more like a commercial product: Things only working through the UI, not through config files (YMMV, of course you still can modify config files, but that’s not supported anymore). Things breaking with every nth update because of arbitrary decisions the developers made.
You know what I do with home assistant updates? Wait at least until mid of month before upgrading to the next major release. And before that very carefully reading the community thread about the release and the summaries of all Github bug reports since the release date to know if still something will break and if yes at least get an idea of the blast radius.
I have been trying to follow logs but they seem to be nonexistent once it fails at the google cast. I find it odd that radio-browser integration works fine in a browser and other integrations (DNLA server) work with google cast. it seems the combination of radio-browser and google cast dont play nice together now. Also strange is I have a full back up of 2024.6.4 and if I roll back to it I am having the same error. I have an image of the sd card before the upgrade and restore (2025.6.4) and it still works. I will lose my device history if I just swap cards with the backup and wait for the next upgrade to try again, so I am still trying to find something that will provide a solution or clues to what is actually broke. at least I can still collect my weather station data and energy data while I wait for a fix to my streaming audio.
I’ve commented before that I think the pace of development is too fast. I’d be ok with fewer new features and more emphasis on stability and new hardware and integrations.
I usually wait until late in the month to try the update, and I always do a full VM backup before I install it. This month I tried upgrading earlier and had problems, so I reverted to the backup. Upgrades are always scary and I don’t like that.
Please give a source for this.
Notice that @crzynik sourced their statement. It only makes sense!
Ummm, yeah. Do that with every release/upgrade to any software! Great tip for everyone.
Define ‘n’ please.
Nope.
I know I have not been here as long as many others. But, for my HA tenur, I have taken EVERY update the day it came out. I believe I have only had an issue ONE time.
But, that is some great advice above. Thanks for your input on how to correctly investigate updates before upgrading.
Same here! My installations are updated to the lastest, but the first install of that system was my third install of HA-OS. One on my old Pi that ran for around two years, and two on my new homeserver. This one is now running for four years something. That move was done via a normal backup and it was all good.
For most people HA runs just fine. People do update in a matter of days after new versions come out, see the statistics.
I’m starting to get the feeling, that the problem might not be HA, but rather between screen and chair…
Is this now again “all against someone [here jeffcrom]” who allowed himself to say something negative?
I didn’t have that many problems in the past, but everyone should at least see, that there are more and more really severe problems in the last releases and months.
Easy to say “your problem”, “custom here and there”, but if so many custom integrations work before (and until the HA core upgrade) and perhaps rely on stable (?) functions and apis, I don’t know if the frequency of breaking changing basics, functions and apis are really the best idea of all, if other integration then break and break HA then as well. Example the changes and problems with the entity history this release.
I have the feeling that HA grew up with the custom cards and integration. And now it breaks more and more with this concept (to get rid of all this?) and don’t care about them any more. And this is not because of “open source” because these changes are only allowed and well approved only from the inner circle of the project.
My installations are updated to the lastest, but the first install of that system was my third install of HA-OS. One on my old Pi that ran for around two years, and two on my new homeserver. This one is now running for four years something. That move was done via a normal backup and it was all good.