Are there too many new features being introduce, too quickly?

Well you convinced me at least. I joined the beta.

2 Likes

Hi everybody,

I’m a bit late to the party, but I wonder whether anybody has proposed more control over the update process? As far as I can see, the only easy option currently is to update to the latest released version. What if one could select the concrete version? This would allow some users to always be on the latest (which is important to discover bugs), others to e.g. update to the latest dot version (the one before the next major release) for improved stability. This would be extra convenient if HA could be configured such that a notification for “update available” would only be shown for that particular version (but it would still be nice to choose that version even after the next major release has been pushed out the door - otherwise, one might miss the rather short time frame during which that version is available, and the next major version has not yet been released).

I really appreciate the pace HA is developed with, but personally would like to update my HA installation once a month “only”, so that solution would suit me just perfectly :slight_smile: - but I guess that there might be quite a few users who would prefer such an approach. I don’t have any idea on how difficult it would be to implement this, though.

Best
Christian

What does “concrete version” mean?

Sorry for not being concrete enough :wink:

What I meant is “let me select from the last (say) 10 releases (which would be 2021.12.7 to 2022.2.5 if I have counted correctly) or all releases of the last (say) 3 months”.

Btw, for me it would already be helpful if the last dot release before a new major version would be marked as such in the release notes only - I’m actually reading this stuff, and I appreciate the work that goes into that!

You can install any version you want. https://www.home-assistant.io/common-tasks/os#usage-examples

2 Likes

Frankly, I didn’t know that - thanks for letting me know! As an IT guy, that solves all my concerns.

But wouldn’t it then be rather easy to support the described functionality via the HA UI to make it easily accessible for anybody?

As I don’t have a test system and my OH is fed up with the issues I have been having when updating I now never install the x.0 version. I wait for a couple of updates to occur before installing and then only if the update has something of interest or impacts a feature I use.
As an example:
2022.2.0 - Nope
2022.2.1 and 2 - Nope
2022.2.3 - installed
This has reduced the amount of issues I experience.

1 Like

That’s exactly the idea here. From my experience, it is rather common in industry to only install the latest service release on critical systems if the according software has a fast release cycle, and if no “long term support” version is available (and if you still need to stay kind of “up-to-date”).

Since HA is not critical to me, but I also don’t want to spend a lot of time maintaining it and stay up-to-date at the same time, that’s the update strategy I will use from now on (that I have learned about the option to install arbitrary versions as described above). I think that this is an awesome compromise. Note, however, if everybody does that, then the services releases will still contain bugs :slight_smile: - early adopters are still needed for that approach…

2 Likes

I run the Beta on my production system without issue, usually. When there is one it takes 5 minutes at most to roll back a version. 15 if it requires a backup restore (hasn’t happened for two years).

Heck, I ran beta versions when I was 5000km from home for a year.

The more people on the beta team the more bugs get caught before release. There is only a small set of testers and we dont run all the integrations.

Having said that I realise it is not for everyone.

1 Like

I applaud you and those who test the beta. Unfortunately I have had many problems with HA and in particular the ZigBee network (which I realise is largely separate from HA.

Are you using ZHA?

If you use Zigbee2MQTT it runs in its own container, separate to Home Assistant.

Originally I was using Deconz in a container but moved to ZHA when I changed to HA OS. Slowly it’s become more stable. Been thinking about Z2M but now thinks are stable don’t want to upset things. In the process of building a new instance of HA on a pine 64 and will try Z2M on that.

With Home Assistant Container (Docker) I find it quite easy to update to a different version. I recently updates to the latest 2021.12.x after 2022.2.0 was released. No issues whatsoever.

My experience from a few years ago and watching the release threads has me using the previous major release. So far, no issues. I tried your idea a few years ago with little success.

1 Like

Funny, because 2022.2.4 was totally screwed and pulled soon after release. Good thing you didn’t wait a little longer and get that one :slight_smile:

Actually the commit that completely killed the MQTT integration in that release was a quick cleanup / refactoring of the debug logging system in the integration that was squeezed in between two dot releases, pretty much without any testing at all. We all saw the results of this.

Dot releases should be only bug fixes, no new features, no unnecessary cleanups or refactorings. These should be reserved for .0 updates.

3 Likes

No you have an idea why I upgrade to the PREVIOUS major patched release :wink:

I let everyone else beta test for me.

Much as I hate to be “that guy,” I’m coming around to this way of thinking. When I did this stuff for work, it was always a good practice to “leapfrog” versions to cut down on the inevitable hassles every update brings. And never install a .0 update.

This 2022.2 release cycle seems like a good example of this. Most of the issues which have come up don’t impact me, but reading about how many there have been makes me think I could be next. I really like to stay current with releases, but it looks like new issues are being introduced with each point update, along with the fixes from the previous one. It’s not clear if or when that will stop.

It’s looking more and more like, if you want your HA implementation to be anything approaching reliable, you need an identical test environment. Same hardware, same sensors, same integrations, everything. Then you could do proper testing, even get involved in beta testing. And if you want to fool around with adding new hardware or features, having three systems starts to make sense: development, test and production. This hobby is starting to get expensive and time-consuming!

Nah. It takes minutes to roll back versions at most, or slightly longer restore a backup (that you should have created before updating).

1 Like

If everyone only installed .4 and up then .4 becomes the unstable version.

I think the automatic creation of a backup before you upgrade and the easy with which you can roll back makes it pretty safe.

I can better understand that people may wait from the Wednesday release to the weekend so you have more time on hand on a day off in case something goes wrong. A normal weekday evening if you have wife, kids and the whole disaster can be stressful.

But advising everyone to ignore the first releases as the ideal goes against the principle of community development. Not all can program Python at the level required on this project. But we can all test and we can all report issues so they can get fixed.

4 Likes