#devs_os-archived

1 messages · Page 1 of 1 (latest)

candid citrus
#

Hi !
I have a Home Assistant OS that can’t update.
I've pinpointed the issue to what I think is an OS broblem / bug : docker can’t pull images.
Docker can’t seem to calculate the image layers’s sha256sum consistently (when it’s big in size) and fail to load any update images.

#

Here’s a screenshot that illustrate the problem :

#

As you can see, when calculated manually, the layer that is supposed to fail the sha256sum, has the expected hash

#

It makes no sense to me

#

All this is happening on a Odroid install, with an OS that has not been tampered with.

#

I have a lot of free space (eMMc of 128 GB, most of it empty) and no proxy (I’ve seen suggestions that a proxy can cause docker some problems)

#

It’s been happening for several months.
homeassistant: 2022.5.5
operating_system: Home Assistant OS 8.4
supervisor: 2022.07.0

#

What I don’t get is that ha info reports « docker: 20.10.14» but on a ssh session, docker -v reports « 20.10.17 »

#

(well ok, I get it now : I was ssh-ing in a container, not the OS)

dusk phoenix
#

@candid citrus are you still experience this? Best is to open an issue.

undone mauve
#

Just some thoughts, why does HA OS use ext4? This is now the second time I've had issues due to a power loss. My other systems that use hash-referenced journals like BTRFS and ZFS don't have this issue since they can discard partially-written files after a sudden interruption. HA OS is basically an appliance designed to come up and down with power, seems like a good idea to use a more modern/improved FS? Or maybe I'm missing something.

dusk phoenix
#

Do your other system experience a power loss at the same rate?

#

ZFS is out-of-tree file system. I am not going to support that for various architectures 😅

#

BTRFS: Sure would be an option. But to me it still seems less battle tested than ext4.

#

Also:

they can discard partially-written files after a sudden interruption
Can you backup that claim?

undone mauve
#

Yeah I would not go for ZFS here unless HA OS was BSD lol

#

Another question is performance. I'm certain BTRFS would be great on a NUC or similar device, but I'm unsure about a pi3 for example

dusk phoenix
#

Sure CoW, has some interesting properties, but afaik its still doesn't guarantee data loss on power failure. ext4 guarantees metadata consistency. All data written continously are databases, which typically have the same guarantees. What kind of problems did you epxierence?

rare sable
#

ZFS and BTRFS (I used more the first one) are nice file system for server with a lot memory, but not good for embedded devices

undone mauve
#

I had issues with mariadb most recently, and a while ago a failure to boot at all (that was a long time ago, so probably some shifting of how HA OS works since then)

dusk phoenix
#

Hm, I did notice that mariadb problems appeared more often lately, there might be a mariadb problem, not sure 😰

rare sable
#

maybe the shutdown timeout is to short and it get killed before he flushed to disk

undone mauve
#

No logs, unfortunately, had to get it working again quickly so I could get my lights working again lol

dusk phoenix
#

That should (tm) not matter as databases maintain a log which should always bring them back up to a consistent state (🙏 ACID). A proper shutdown commits all the logs, and makes a restart faster/not requiring a log rollback, but still, a power loss should not cause a database corruption (in absence of hardware failure... Of course, if HW does not behave according to spec (e.g. write ordering) all thigns are lost.

#

But in face of broken HW all file systems will fail (except /dev/null of course 😄 )

undone mauve
#

yeah ok I'll double check this ssd

dusk phoenix
undone mauve
#

Yeah, though I am curious about the performance of BTRFS on a pi4. Synology is using it in their Celeron-based devices, and that's an actual RAID array with many simultaneous operations

#

I'm struggling to find any benchmarks for a Pi4 w/ BTRFS so who knows 🤷

rare sable
#

BTRFS and ZFS are always slow as a direct native file system BUT they have a lot of software logic inside that allows building fast and quite nice SAN systems with raid write/read cache and using memory. That stuff is nice with all the features, but standalone 🤷‍♂️

#

allow them to use 1-8GB memory for read cache and that stuff is like a rocket on the read IOs

#

but as focus to embedded devices, memory is all time what we not have

dusk phoenix
#

In the end MariaDB itself guarantees no data loss (in absense of hardware failure and corner cases like disk full, hm is that maybe what has happened 🤔 ). Whatever caused issues here should be tracked and fixed.

Databases open a huge file and do random writes in it, that is actually a workload where CoW file system cause a lot of fragmentation. Fragmented in theory can lead to write amplication effects on flash devices. In fact, MariaDB recommends to disable CoW for its datafiles:
https://mariadb.com/resources/blog/what-is-the-best-linux-filesystem-for-mariadb/

It’s usually best to mount Btrfs with the nodatacow option, disabling copy-on-write, because COW causes fragmentation, dish thrashing, and CPU and RAM spikes when you have a lot of random writes.

undone mauve
#

I unfortunately don't have logs, but I will be sure to extract them next time

candid citrus
knotty basin
#

Hi! I want to import some configuration from an USB stick for an USB stick. I found this documentation https://github.com/home-assistant/operating-system/blob/dev/Documentation/configuration.md

But I don't know if I put my udev rule directly into the udev folder or do I need the rules.d subfolder inside this? Also the mentioned command systemctl restart hassos-config could not be found on my HA when I login with ssh. Is this the same as ha os import?

dusk phoenix
#

Yes, those two are essentially the same

knotty basin
#

I did not get the "[Info] Update Udev configuration!"

dusk phoenix
#

systemctl status hassos-config

knotty basin
silk flame
#

Guess it shouldn't take an hour (by now) to restart my pi after 8.5 update? 🤣

vivid fractal
#

Hello, how can I add firmware file for my BT dongle (LM1010-0972) to HAOS?
[ 3.423238] Bluetooth: hci0: BCM20702A1 (001.002.014) build 0000
[ 3.427814] Bluetooth: hci0: BCM: firmware Patch file not found, tried:
[ 3.427849] Bluetooth: hci0: BCM: 'brcm/BCM20702A1-0a5c-21e8.hcd'
[ 3.427860] Bluetooth: hci0: BCM: 'brcm/BCM-0a5c-21e8.hcd'
Thanks

  • except getting operating-system from git, adding the firmware patch there and all the rest?
inner widget
#

9.0.rc1 release action failed, manually updated using terminal (ova) and seems all good.

hollow creek
dusk phoenix
dusk phoenix
# hollow creek do we have some relase notes?

Not yet, I need to put it toghether. It is not that big of a change as the Buildroot version (and with that most packages like systemd, Docker etc) and kernel version stayed the same this time around.

#

Probably the biggest change is the switch to cgroups v2

hollow creek
#

upgrade went good through

dusk phoenix
bleak cairn
#

Just realized this forum exists... I posted this on the beta, but I updated to 9.0 Rc1 and it's been about 25 minutes and my HA instance has not come back yet. SSD drive is still flashing. Tell me it's just taking its time because it's a big update? 😦

#

should I just turn off the system and turn it back on? its been over 30 minutes now

#

ok, after about 40 mins i just manually turned off the RPi and its back up and running. Hate doing that, but its not often that I do. But its updated now: Home Assistant 2022.9.0b3
Supervisor 2022.08.6
Operating System 9.0.rc1
Frontend 20220902.0 - latest

inner widget
bleak cairn
bleak cairn
#

hey just curious is RC version considered beta?

rare sable
#

do you use the usb disk feature or the unsupported usb boot?

#

If you use the last one, you should reinstall the system and using usb disk, we tell people now over years that usb boot is bugy and only a sell driver for the raspberry pi with an very uncommon praxis in the IT space. No one buy an computer in a shop to boot later from a usb disk instead over the internal harddisk

fallow scarab
#

I see the USB/IP kernel modules in the HA OS 9RC. I have a spare RPI that I installed USB/IP on. Before I just dive in, is there any other information on testing this?

rare sable
#

It need an addon or someone adding support to supervisor with API that it can make an UI around

placid marsh
#

how often does the buildroot repo get synched with upstream and modified? or is the same version used for every release

#

it doesnt show up as forked on home-assistant/buildroot

dusk phoenix
#

Yeah its not a forked repo repository in GH sense, but it is the same repository as in "it has common history" We are actively using the 2022.02.x-haos branch only. There are a bunch of patches ontop of vanilla upstream 2022.02, you can list them by looking at something like git log 2022.02.5..2022.02.x-haos. You have to have the upstream buildroot repo as another remote to get all the latest tags.

dusk phoenix
placid marsh
placid marsh
#

did you guys use the menuconfig tui for the initial settings, and then modified it as needed, or was it all custom config from scratch for the packages and kernel params?

dusk phoenix
#

Not sure, the initial buildroot defconfigs are very old at this point. The kernel configs depend on the board, but I think we made all of them from custom defconfigs

rare sable
#

Yes, I did all from scratch as our OS specified what is needed not the example defconfigs

bleak cairn
#

If you use the last one you should

sinful parcel
#

I was looking at the most recent iteration of the platform host and am wondering why you ended up putting the kernel in its own squashfs/partition rather than putting it into the system partition. /lib/modules still has to stay in sync so I'm guessing its not to make debugging easier, is there a limitation in uboot that the kernel has to be by itself in a partition/filesystem?

sinful parcel
#

(maybe this is a subtelty with RAUC, though I don't see any requirement in their docs that a slot is broken into 2 physical partitions)

placid marsh
#

is there a source file for kernel compile params that is pulled by all boards/configurations? So e.g. I want to add a kernel compile flag for all supported platforms. Or would I have to add it in each respective kernel.config?

#

looks like device-support.config hassos.config, and docker.config get included in the different boards and are common hmm

dusk phoenix
sinful parcel
#

Thanks for your insight!

dusk phoenix
dusk phoenix
#

I guess at one point we need to repartition... It will be major pain, if possible at all 😰 Maybe we should consider consolidating then.

placid marsh
#

@dusk phoenix when you guys make changes to the kernel config between releases, do you adjust the kernel configs directly, or do you use menuconfig?

#

and sorry for atting/bugging - I'm looking to make some contributions, but I want to understand the team's workflow and the general architecture at a level that will make my contributions easier to integrate (if accepted)

rare sable
#

I split that up. We have board kernel configs that is needed to get the board running. Then we have docker kernel options that is needed to get docker working and then we have our own kernel configs (haos) that we need to have the same interface to the userland and we have some devices that we have to support but are not relevant for board/haos or docker

#

based on what you want modify, you can but it into correct kernel config. And no, I don't use menuconfig at all

#

so the board kernel config is the base, and on top of that we apply our global kernel settings and make sure they are set as we want to make haos working same over all boards

#

@dusk phoenix but yeah, instead 3 files, it could be also one file with comments on it to know what is needed by what component

#

At least at my time on the OS, I never had major issues with that

dusk phoenix
#

I use menuconfig in my kernel tree separatly, usually to better understand dependencies. Sometimes they can get tricky.

#

But once I know what symbols I want to change, I just copy those to the config files, usually roughly in order of the kernel .config. But it doesn't matter where the configs are, but it makes sense to group them roughly if they are grouped in regular .config.

placid marsh
#

@rare sable I don't mind the kernel config being separated into files. menuconfig does make it a lot easier to deal with dependencies, and I find it to be quicker than manually digging through the necessary params.

I was thinking of write a quick shell script to concatenate the various kernel params that HAOS uses for the platform/board I'm interested in (currently ova/amd64). Then I can make a copy and name it .config, run menuconfig on that, and visually modify the things I want to modify. I am then left with a single updated file that I can diff with the original, and then I can decide which file in the haos source tree it should go into.

I fixed a hotplug issue the other day that I want to submit a PR for, but I did it without menuconfig, and I just wanted to be sure that I didn't miss any dep and that I didn't add things that are not necessary, and menuconfig would act as a quick second set of eyes so to say. So I guess this workflow is somewhat similar to what @dusk phoenix's

While researching the fix, I also realized that there were a couple other related items I wanted to add for better hypervisor support. Not sure if I should submit everything as a generic "improve hypervisor ....." or do a separate PR so that each one has a more narrow scope.

#

Or would it be better if I created a github issue for this with my proposed fix, get everything ironed out, and create a PR based on the input I get then?

#

I'm heading out for the night, but I will keep an eye on this discussion. Appreciate your input thus far. The past few weeks have been a learning experience, and I have grown to like this project quite a lot.

dusk phoenix
#

While researching the fix, I also realized that there were a couple other related items I wanted to add for better hypervisor support. Not sure if I should submit everything as a generic "improve hypervisor ....." or do a separate PR so that each one has a more narrow scope.
@placid marsh generic PR is fine. Don't overthink it, we have very few contributors to the OS and don't have very strict policies in place when it comes to contribution.

#

An issue make sense if you have not done any work, and like to have conceptional input. But it sounds you are already in code. I think in this case a draft PR is better. Then we can discuss the actual changes easily. And we can also close it and open new/separate PRs if necessary.

#

I fixed a hotplug issue the other day that I want to submit a PR for, but I did it without menuconfig, and I just wanted to be sure that I didn't miss any dep and that I didn't add things that are not necessary, and menuconfig would act as a quick second set of eyes so to say. So I guess this workflow is somewhat similar to what @dusk phoenix's
In cases I am no sure if I got the deps right I check the .config after a build. Usually a quick thing, and then you are sure you really get what you expect 🙂

placid marsh
dusk phoenix
#

Good work! I can give it a spin on my Virt-Manager/libvirt/qemu setup.

placid marsh
dusk phoenix
#

That specific line

placid marsh
#

I'm not really sure which line

#

the message sits between two edits

dusk phoenix
#

CONFIG_VIRTIO_MMIO_CMDLINE_DEVICES=y

#

The comments refer to the line above it 🙂

#

Always 🙂

placid marsh
#

Ah, gotcha, and yeah, I would have to review my notes to see why I added that in. I got up because I had a hard time sleeping, so figured I'd check the computer, and noticed the PR had an update, but I'm about to crash again heh 🙂
I can have a look tomorrow and get back to you, but iirc that line was a tangential thing

placid marsh
#

hm yeah i think i added it in for flexibility because mmio had already been set to y, and iirc openwrt had it set for their appliance too. but it's for rare legacy stuff afaik

#

time to pass out 🙂

dusk phoenix
#

Some virtualization methods allow to load a Linux kernel directly (qemu), in that case the VM monitor can pass kernel command line parameters directly. So in this world, qemu can control mmio device. However, that won't work with OVA as that is a full image with boot loader, there is no direct control of the kernel command line from the VMM. However, we have cmdline.txt and the option to manually edit the kernel command line in the boot loader. So for testing this might come in handy. I don't mind that is enabled, just wanted to point out.

placid marsh
# dusk phoenix Some virtualization methods allow to load a Linux kernel directly (qemu), in tha...

Thank you so much for explaining. When I said legacy, I was referring to mmio in general (4.2: https://docs.oasis-open.org/virtio/virtio/v1.1/csprd01/virtio-v1.1-csprd01.html#x1-1440002) which I thought was the legacy way of interfacing with devices. I do like that qemu allows the loading of a kernel directly and supplying a separate rootfs. I agree that having the cmdline.txt load params options is handy.
There are a number of debug-ey options that I wouldn't mind having (not just in the ova, but the non-virt images too). I was going to suggest that we create a separate debug version, with a number of additions for troubleshooting, and a number of restrictions loosened up. This would be marked as debug use only (though I do recognize that this would likely tempt some people). For example, I was going to submit a PR for enabling /proc/config.gz and a few corresponding userspace bins (zgrep, zcat). I don't think that this would inflate the image too much, but it's also not something that the average appliance user would want/need to use.
Have you considered creating a debug build for appliance devs?

dusk phoenix
#

Not really. I am down to add config.gz though, I don't think it will cost much

placid marsh
# dusk phoenix Not really. I am down to add config.gz though, I don't think it will cost much

hm what does your troubleshooting workflow look like when it comes to bare metal? do you rely on ssh/p22222 as far as hands on goes?

There is some device support I wouldn't mind having a go at adding, and knowing how the core team does things would be very helpful 🙂 I have a bit of a laundry list of ideas for HA and HAOS.

I can submit a PR for config.gz and parsing tools 🙂 (busybox provides very basic stream decompression tools, but not zgrep afaik. buildroot can provide zgrep. let me know if you have a preference on what to add). I was thinking this could be handy not just for the ova, but also the bare metal/sbc releases.

dusk phoenix
#

We can use zcat /proc/config.gz | grep, zcat is available from busybox.

#

Most of the time SSH on port 22222 yes. For ARM boards, I often use serial console as well.

#

On ova there is also a serial console, sometimes helpful there as well. I use systemd.unit=rescue.target as cmdline sometimes to debug systemd units/dependencies stuff. It gives you a shell in a "half booted system", ideal to poke around and start services and see what exactly gets pulled in with it.

#

What I was thinking is to create a debug add-on, essentially an addon with tools to inspect the system. An addon which runs privileged etc is basically a shell on bare metal 😄

placid marsh
placid marsh
#

any chance you could modify the build script/workflow so that the dirs and files that are created aren't owned by root, but by the local user (or maybe a group)? I keep having to use the terminal to delete stuff and it's a bit tedious (I like the convenience of the file manager for some things)

rare sable
#

The container which built the system can handle it

placid marsh
rare sable
#

the script that enter into our build backend

placid marsh
#

are you suggesting i modify it to fix perms on my end, or PR it upstream? 🙂

rare sable
#

if you use that, you should not have permission issue

digital quail
placid marsh
# rare sable if you use that, you should not have permission issue

I use the build scripts, which in turn use the container. Everything that the script generates under operating-system/release is created as owned by root. Also every file under output*/images/ is owned by root:root
When I build, it looks something like this: sudo scripts/enter.sh make O=output_rpi3_64 rpi3_64

placid marsh
#

Has anyone been able to run the rpi3 aarch64 images in qemu for testing? I am able to get it to boot, but it hits this stage and then unloads and gracefully shuts down:

[FAILED] Failed to start Docker Application Container Engine.
See 'systemctl status docker.service' for details.
[DEPEND] Dependency failed for HassOS supervisor.

I've added a missing (I think) kernel module for the USB network adapter thinking that it was failing and shutting down because of missing networking, as time sync also fails, but I later tried the ova build on x86 with the network disconnected and it still booted fully. So I'm not sure where to go from here. I am hoping to be able to run it emulated to test a few things. I do have a physical pi3, but I really want to get it to work emulated. I imagine that this could be useful to you as well, @rare sable , if you don't already have an emulated setup.

placid marsh
placid marsh
#

I dumped a raspbian image in place of the haos one (and the corresponding kernel and device table files), and networking and time sync works in it. Not sure if this matters, but at least this verifies that my tap+bridge config is working

#

I am going to see if i can get docker to install and run on the raspbian image, just to rule out qemu being the problem

placid marsh
#

@rare sable can I DM you super quick with a question on the topic of security (in relation to haos)?

dusk phoenix
#

@placid marsh sorry, I was on leave. Back in the "office", now 😅 Feel free to DM me.

dusk phoenix
tepid hollow
#

My raspbian rpi SD died, so I copied /config and created a VM of HA OS, which is so much nicer and easier to manage. Now need to figure out how to boot HA OS off a USB drive so I don't need to fix the SD. Any special sauce?

placid marsh
placid marsh
# dusk phoenix So far never used Qemu + Raspberry Pi 3 machine to run the RPi3 image via emulat...

I had to leave it at that for the time being, unfortunately. But I'm still hoping to get it to work.
I was loading uboot as the kernel, so I don't think that that part was/is the issue. If and when I get this to work, I will post an update. It could be handy for some forms of testing before testing on actual hardware. Though even if no one else finds it handy, it will be a good learning exp.

rare sable
#

Yeah, we have hardware that can flash an RPI remote

languid fern
#

Anyone use Gitlab for the development of OS? I'm looking to make some minor changes and wanted to see if there was an existing .gitlab-ci.yml file that people were using that hasn't been included in the repo?

full brook
#

Hi. If I wanted to raise a PR to sync package(s) to the latest upstream buildroot, do I make the changes in homeassistant/os? Can I simply fetch the upstream branch, and cherry-pick or merge the changes into os, test, then raise the PR from there?

#

In this case I'd like to take bluez5-utils (and -header) up to 5.65

dusk phoenix
#

We have a separate repo for buildroot itself (its a submodule in the operating-system). You can try to cherry pick from buildroot master and create a pr, however, we do use the last LTS which is 7 months old at this point

#

Any specific reason you'd like to update bluez?

full brook
# dusk phoenix Any specific reason you'd like to update bluez?

I'm having issues connecting to a BT audio receiver within a VM. I was able to pair and trust the device successfully using bluetoothctl. On reboot, the device will not autoconnect (it attempts to connect indefinitely after power on). When I try to connect manually I get:

Attempting to connect to E3:28:E9:23:19:10
[CHG] Device E3:28:E9:23:19:10 Connected: yes
Failed to connect: org.bluez.Error.Failed br-connection-unknown

Logs for bluetoothd have:

Oct 05 21:12:52 homeassistant bluetoothd[346]: profiles/audio/avdtp.c:avdtp_connect_cb() connect to E3:28:E9:23:19:10: Permission denied (13)
Oct 05 21:12:52 homeassistant bluetoothd[346]: src/service.c:btd_service_connect() a2dp-source profile connect failed for E3:28:E9:23:19:10: Device or resource busy
Oct 05 21:12:59 homeassistant bluetoothd[346]: profiles/audio/avdtp.c:avdtp_connect_cb() connect to E3:28:E9:23:19:10: Permission denied (13)

I figured since bluez hasn't been updated in awhile and a lot of protocol bugs have been fixed btw 5.63 - 5.65, it would be good for BT support across the board. I wanted to be sure there weren't any reasons to stay back and that cherry-picking from BR master was the way to go

#

I was just about to enable debug logging on bluetoothd to identify a specific error

tepid hollow
#

What's boot doing when it says:
22-10-13 13:11:06 INFO (MainThread) [supervisor.resolution.fixup] System autofix complete
Seems like my new HAOS9.2 install is stuck there. Waiting for images to pull in another thread?

dusk phoenix
distant sorrel
#

If I am running the X86 image provided, is there any way to get dmidecode or lshw?

dusk phoenix
#

@distant sorrel sorry for the delay. There is no package manager on HAOS, so on the HAOS shell, no. But the homeassistant container runs privileged, so temporarily you can just exec into it and use the Alpine package manager to add stuff 😅 (docker exec -it homeassistant /bin/bash).

sinful mantle
#

Currently running OS 9.2, and getting a notification to upgrade to 9.3. Pressed install and get an error saying another install is in process, but it's been running for over 24 hr's. I've already rebooted a few times. Is this a known issue? Would appreciate any thoughts on how to clear it, or if that might damage things

latent wigeon
haughty path
#

hello, is there a changelog available for the dev versions of HAOS?

inner widget
#

Yes, the commit history 😉

haughty path
#

ok thank you

#

the reason I ask is to see if there are any changes in the aarch64 generic dev image that would enable additional sbcs to work correctly (orange pi 4 lts)

dusk phoenix
drowsy shell
#

I'm so in over my head. "HAOS for rock5b" Im attempting to use buildroot to MAKE haos for rock5b. So far I can't clone the buildroot.rockchip.ext into the hassio buildroot-external dir. I was able to clone to self named sit and then cp -r -n /that/dir/* /this/dir/ but still can't make for the rock5b.. any help would be awesome. Obviously on a different thread or PM. Sorry again for posting here....

velvet sierra
#

Support Of New Hardware

light totem
#

Would be really nice if the OS wouldn't hang waiting for the root partition in Hyper-V

copper agate
#

Do I understand this correctly that HAOS currently cannot be built on an arm64 host?

>>> host-go-bootstrap 1.4-bootstrap-20171003 Building
go tool dist: unknown architecture: aarch64
copper agate
#

Do I understand this correctly that HAOS

upper nebula
#

Hello Team

upper nebula
#

it got fixed after few reboots.

clever basin
#

Hi everyone. I have a Bluetooth device that I'm trying to integrate with HA, but BlueZ 5.63, the version included with HA OS, has a bug that prevents it from working properly (see https://github.com/bluez/bluez/issues/438). The bug has now been fixed in the latest master branch of BlueZ, but not yet included in a release of BlueZ. Is there a way that HA OS can cherry-pick the two commits that fix the bug, and apply the patch to the included version of BlueZ?

These are the commits that would need to be cherry-picked:
https://github.com/bluez/bluez/commit/b91d9213d951277896845830d0680061c7643828
https://github.com/bluez/bluez/commit/da203f5dbc7edaa15b55d1efcea3fda5e43cfc2e

clever basin
raw storm
#

Hi, I run HA (VMware ESXi/vSphere (.ova) in Synology Virtual Machine Manager. HA VM doesn't want to start anymore, probably because I have some files with checksum mismatch on my system volume. I was able to create a new HA VM and would like to try to restore from a backup from my old VM. IS there a way to find those backups even if the VM doesn't want to start? I do have root access to my Synology system volume. Thanks for your help

maiden turtle
#

Hi, sorry for bothering here but I don't know better, so please forgive me if I may be in the wrong place.

I am running HAOS on an ODROID H3.
My issue is the default HAOS image seems to not have the firmware files integrated for the RTL NIC that is on that board.
I have the fw files that would be needed for running the NIC correctly but I do not know a way to integrate them.
I found out how to access the root shell of HAOS and can also mount my USB stick that contains the fw files.
The only thing missing is: how do I get the fw Files to /lib/firmware/* ?

dusk phoenix
#

The root file system uses a read-only squashfs, so integrating is not entirly trivial. There are people which managed to extract/rebuild the squashfs, but that is not the recommended process.

The right way is to integrate the firmware files to the Buildroot based build system at https://github.com/home-assistant/operating-system/

maiden turtle
#

OK, so I would need to setup a build chain to build a new OS image containing the fw files

dusk phoenix
#

Essentially yes. You can also create a PR so it will be in the next release. What board are you using on the H3?

#

I am guessing generic-x86-64

maiden turtle
#

yes, are there alternatives?

dusk phoenix
#

Not really 🙂 I was just not aware that this is x86 😄

maiden turtle
#

it is an intel N5105

#

could it be possible that the 10 dev version already has the fw files for newer intel chips already integrated?

maiden turtle
#

does HAOS not have an ntfs driver for mounting?

dusk phoenix
#

Could be it is not enabled (or maybe not on all platform). I am ok enabling it by default.

#

The whole mounting is something which needs more work. We don't have a recommended way to mount external drives or network locations 😰

maiden turtle
#

well i like the udev rule 80-...

#

if this would be a bit more accessible it would be great, as it can be customised

#

i would also suggest, if possible, to change the file system layout a little bit. E.g. adding new firmware files always includes recompiling via build root. this is very impractical

#

if e.g. /lib/firmware would be mounted as a rw partition from disk it would be so much easier

#

yes, it would require changing the system image layout a bit and reserving a space which was before in the squashfs, but it would make life so much easier as you do not need the whole recompile thing. and also you would not have to wait for new OS releases

#

getting back to the mounting: i am kinda ok with the automounting as /media/xxxx because it makes the disk accessible through the UI on a minimum level. with the addons like samba-nas you can have more access if needed, so thats fine for me at least

tepid hollow
#

I was coming here to ask about an issue, but I see Florian is already asking about it.

#

Would an overlay filesystem (for /lib/firmware) be a workable solution on top of squashfs?

vast palm
#

While it’s inconvenient for the first user to report the issue, I think getting the firmware added to HAOS going forward is still the better option. So users that might not be comfortable poking firmware benefit, If people require more control they might be better running the container directly on a mutable OS base.

dusk phoenix
#

In general, we ship the firmware maintained by the upstream Linux kernel devs. Usually we bump to the latest versions whenever a new kernel gets shipped. Firmwares are sometimes very tied to the kernel driver (version). So IMHO, most of the time it is fine that firmware are shipped as part of the rootfs with everything else.

That said, there are sometimes corner cases where it would be nice if users could add firmware themselfs, e.g. for testing or if the license of a firmware prohibits redistribution by us.

Besides /usr/lib/firmware the kernel actually browses a second path for firmware, with higher priority at /usr/lib/firmware/updates. I suggest to symlink that to a writeable location, proposed here:
https://github.com/home-assistant/operating-system/pull/2225

velvet sierra
#

Hi, sorry to bother you, but I really don't know where to ask. Please forgive me if I'm wrong here.

I running HAOS 9.4 and wanted to use the GStreamer integration for HA. But in order to be able to use the GStreamer integration some packages have to be installed on the system first. (as for ffmpeg for example)
Unfortunately these packages seem to be missing from HAOS, so I currently install them in the HA container after every update, which is a bit of a hassle. I would really appreciate some help on this!

More context: https://github.com/home-assistant/core/issues/72875
(Issue seems wrong on the "core" project, but here it started.)

viscid thicketBOT
#
elektr0nisch#6741

@wild halo

Joined

Wed, Dec 21, 2022 10:25 PM

Registered

Mon, Sep 11, 2017 3:37 PM

Roles [0]

None

ID: 356825466855751691 ·
hasty pagoda
# viscid thicket

😅 whoops, was trying to find the person who commented on my GitHub issue for this channel, it seemed to autocomplete to the last poster in the channel

hasty pagoda
#

Is anybody in here running Home

plain steppe
maiden turtle
plain steppe
#

Oh with 5.15 it's built in anyway, so just needs to be enabled with CONFIG_NTFS3_FS=y

plain steppe
#

Im having a weird issue with post-build

#

tempio seems to hang in readConfig

maiden turtle
# plain steppe Oh with 5.15 it's built in anyway, so just needs to be enabled with `CONFIG_NTFS...

Yeah but this would mean: installing a build chain, compiling an Image, installing that, setting up everything. Again and again and again.
For nothing else than built in ntfs support. Without automount of externally plugged in disks. And my problem with missing fw files still would exìst.
Sorry, but it sounds a bit counter intuitive to me and kind of makes the concept of HAOS superficial. There are conceptual issues that should be fixed first in my opinion.

plain steppe
#

unclear why there's a sub shell involved at all

pastel geyser
#

I've just moved from HA running as a container on a large linux box to HA Yellow, and I'm trying to understand the layers described here: https://developers.home-assistant.io/docs/architecture_index . Specifically, can I just drop a binary on the core's fs and add a cronjob there, or will that get overwritten next time supervisor updates the core?

rare sable
pastel salmon
#

I'm not familiar with the Yellow but if it's like my Blue, HAOS is the operating system, the docker is HA Supervisor, and HA Core and other add-ons are containers...

dusk phoenix
dusk phoenix
plain steppe
#

seems like it should not need to be inside a sub shell either way

dusk phoenix
dusk phoenix
plain steppe
#

Yea it must be a different shell. I'll check later

copper agate
#

I added support for the Rock Pi 4B+ to HAOS: https://github.com/citruz/haos-rockpi
Please let me know if anyone here wants to test it or help with adding support for more boards if you have another board of the Rock Pi family 🚀

copper agate
#

the 4SE seems to be very similar to the 4B, so it might actually work. Please try and let me know

copper agate
#

Hmm sounds like docker did not come up correctly. Can you check journalctl -u docker?

#

you may need to press alt+right arrow to get to the console

#

journalctl -u docker.socket?

#

any other errors in journalctl?

#

Yeah I was about to say, something about your SD card does not seem to be right since it cannot find the hassos-data partition. The documentation suggests at least 32GB

#

Maybe 16GB is also fine, but 4GB is very tight 😄

#

uhm no I don't think so

#

you could do some custom setup where only the bootloader is on the sd card and the main OS on the USB, but it will be painful

#

you would need to recompile the image yourself with a custom bootloader configuration

#

in theory yes, but I wasn't able to test NVMe support yet

hasty pagoda
tiny arrow
hasty pagoda
#

@tiny arrow thanks, I'll ask again there. I was also trying to get some signal on the GitHub issue, since it seems like there might be a configuration issue with the OS image for this particular hardware (from user reports it seems to not be specific to this hardware though). I'm happy to dig deeper and help resolve the issue, but am looking for guidance from the development team

robust python
#

Hey Dev team, really random question for you 🙂 I've got the HomeAssistantOS installed in a container on my network and I see it appear in Windows -> Network -> "Other Devices" as a clickable device that when you click on it it instantly opens the HAOS admin panel web page. I would love to re-implement this same functionality on some other VMs I have on my network. Can you tell me what package / service you used to set this up?

honest comet
#

Does anyone know if there's a way to add qemu-guest-agent to HAOS?

honest comet
#

Does anyone know if there s a way to add

vast palm
#

is CONFIG_IPV6_ROUTE_INFO enabled on all HAOS flavours? anyway to check? trying to help a user with thread and they seem to be missing the accept_ra_rt_info_max_plen sysctl (have checked and triple checked they are running on the OS and not a container, of course i haven't checked myself so theres still a possibility)

#

They are running 9.5 / qemux86-64

dusk phoenix
#

Usually I check by looking at .config in my build dir. It seems that CONFIG_IPV6_ROUTE_INFO is dependent on CONFIG_IPV6_ROUTER_PREF, which is not set 🤔

#

My Arch kernel has all three of those:

CONFIG_IPV6_ROUTER_PREF=y
CONFIG_IPV6_ROUTE_INFO=y
CONFIG_IPV6_OPTIMISTIC_DAD=y
#

@vast palm should we enable those?

vast palm
#

i think so

dusk phoenix
#

Hm 🤔

vast palm
#

not sure what behaviour is without the accept_ra_rt_info_max_plen, but if the effect is that its 0 then thread (at least with apple BR's) doesn't seem to work

dusk phoenix
vast palm
#

yeah thats what i feared

#

can see that it physically gets to eth0 but no sign of anything in route tables, so suspected that might be the prob

dusk phoenix
#

On my x86-64 machine (which has this disabled) I do see routes from my other OTBR though

#

Wonder if the otbr agent is learning those in user space 😅

vast palm
#

i know my debian box needs accept_ra_rt_info_max_plen set to get my Apple BR's working. does OTBR use route info option (24)? i seem to remember that varied somehow?

dusk phoenix
#

Yes looks like

image.png
vast palm
#

need to leave office but this is what i see:

# 
    hop limit 0, Flags [none], pref medium, router lifetime 0s, reachable time 0ms, retrans timer 0ms
      source link-address option (1), length 8 (1): 04:99:b9:64:1b:eb
        0x0000:  0499 b964 1beb
      prefix info option (3), length 32 (4): fd69:3e50:e165:4bf8::/64, Flags [onlink, auto], valid time 1800s, pref. time 1800s
        0x0000:  40c0 0000 0708 0000 0708 0000 0000 fd69
        0x0010:  3e50 e165 4bf8 0000 0000 0000 0000
      route info option (24), length 16 (2):  fd59:86c6:e5a5::/64, pref=medium, lifetime=1800s
        0x0000:  4000 0000 0708 fd59 86c6 e5a5 0000
#

and they have

# 
    hop limit 0, Flags [none], pref medium, router lifetime 0s, reachable time 0ms, retrans timer 0ms
      source link-address option (1), length 8 (1): a8:51:ab:90:6e:d0
        0x0000:  a851 ab90 6ed0
      prefix info option (3), length 32 (4): fdbe:e98b:173f:4f12::/64, Flags [onlink, auto], valid time 1800s, pref. time 1800s
        0x0000:  40c0 0000 0708 0000 0708 0000 0000 fdbe
        0x0010:  e98b 173f 4f12 0000 0000 0000 0000
      route info option (24), length 16 (2):  fdcd:7fc0:8284::/64, pref=medium, lifetime=1800s
        0x0000:  4000 0000 0708 fdcd 7fc0 8284 0000
      route info option (24), length 16 (2):  fd21:1896:ac15::/64, pref=medium, lifetime=1800s
        0x0000:  4000 0000 0708 fd21 1896 ac15 0000
dusk phoenix
#

They have two prefixes they claim to route, but other than that looks the same to me?

vast palm
#

Yep.

#

So I don’t know why it works for you

dusk phoenix
#

From OTBR logs:

otbr-agent[282]: 06:21:46.999 [I] BorderRouter--: Processing RIO (fd84:da21:5e08:1::/64, 1800 seconds)
otbr-agent[282]: 06:21:46.999 [I] NetDataPublshr: Publishing ExternalRoute fd84:da21:5e08:1::/64
otbr-agent[282]: 06:21:47.000 [I] NetDataPublshr: ExternalRoute fd84:da21:5e08:1::/64 (state:Added) in netdata - total:1, preferred:0, desired:10
vast palm
#

So I read that as it imports the route from userspace like bgp would do, is that what you think it’s doing?

dusk phoenix
#

Yeah, maybe.

vast palm
#

In which case we do want the config option for inter-op with other BRs?

dusk phoenix
#

Yeah true, for systems which do not run a BR.

dusk phoenix
#

I vaguely remember that back a few months when I first run BRs on HAOS I was a bit surprised that sysctl configs for accept_ra_rt_info_max_plen was not necessary there. Now it all kinda make sense.

vast palm
#

yeah, i talked to some people running otbr with homekit_controller and was baffled they'd not had to set that

dusk phoenix
#

I did some more testing with a generic-x86-64 machine, somehow it seems that it's picking up the routes.

#

I'd go for 64 since that is what all Thread BR are using and the OpenThread docs are recommending.

#

The question is which interfaces we are setting it.

#

The setting exists interface specfici, and all variant, however, I somehow don't get how all works (when reading source). To me it seems it can't work.

zinc canopy
#

hello! i am new to development. i am trying to build custom component for Home Assistant OS, Raspberry PI 4. Requirements are custom prepared package (has no issues) and scikit-learn. Once I specify scikit-learn in manifest.json requirements, and I want to restart HA, I get such error: ``Unable to install package scikit-learn==1.1: error: subprocess-exited-with-error × pip subprocess to install build dependencies did not run successfully. │ exit code: 1 ╰─> [79 lines of output] Looking in links: https://wheels.home-assistant.io/musllinux/ Collecting setuptools<60.0 Using cached setuptools-59.8.0-py3-none-any.whl (952 kB) Collecting wheel Using cached wheel-0.38.4-py3-none-any.whl (36 kB) Collecting Cython>=0.28.5 Using cached Cython-0.29.33-py2.py3-none-any.whl (987 kB) Collecting oldest-supported-numpy Using cached oldest_supported_numpy-2022.11.19-py3-none-any.whl (4.9 kB) Collecting scipy>=1.3.2 Using cached scipy-1.10.0.tar.gz (42.4 MB). `In VS code terminal within HA I can install scikit-learn but within requirements it does not work and therefore custom component cannot be used. Is there any solution how to install scikit-learn?

dusk phoenix
#

This is about Home Assistant OS development itself (low level OS Buildroot/Linux). Custom components is a concept of Home Assistant Core, so your question probably belongs to #devs_core-archived

zinc canopy
#

Thank you, will try my luck there

vast palm
vast palm
#

Oh yeah I meant to say 64 is what we’ve been telling users to use so far and it’s worked fine, so all good from my pov if you go with that as the default 👍

dusk phoenix
#

Ok, yeah I need to dig into this a bit more

dusk phoenix
deep cloak
dusk phoenix
#

@vast palm transferred the discussion in a bit more formal place: https://github.com/home-assistant/operating-system/discussions/2333

Seems NM is processing those too. it think for now I'd leave the kernel configs out as we seem to process what is necessary in user space. I wonder how systems behave if both, user space and kernel space process these RA's.

vast palm
#

The question i have is why is NM not processing the RAs in this case. Ipv6 is enabled enough to get an fe80 address. If ip6 was disabled in the haos settings, would it still leave the link local addresses in play? I didn’t check their haos settings because I saw link local addresses and thought it was enough,

#

(Will ask them to check)

dusk phoenix
#

If ip6 was disabled in the haos settings, would it still leave the link local addresses in play?
Yes it does. I am not sure how it behaves in terms of IPv6 Neighboorhood Discovery protocol.

rapid compass
#

Does anyone know any app creator for One UI 4.5? I would like to make an application that would be used to control devices in home assistant. For example, a garage door, etc. At the moment, the widget with quick entity control is poorly solved. Unless he improves it in a few days/weeks. To make it work better and faster, and show the status of devices.

silk tusk
#

noob question... if i want to edit the contents of a single core python file in the homeassistant components .. for testing: can i edit a python file somewhere directly? does it have to somehow compile the python files for that to work? where are the core/homeassistant/components files on hass os?

latent wigeon
#

None of that is related to OS development

#

You can edit any Python file and it will persist until you update or otherwise rebuild the HA container

silk tusk
#

appologies, was thinking #os would be able to identify where the core .py files were...

latent wigeon
#

No

#

They're in the HA container

silk tusk
#

ahh, so hass os is still running a container? like docker? of course!

latent wigeon
#

Yes

silk tusk
#

thanks RobC. 🙂

nocturne stag
#

Something about HAOS is making NetworkManager lose track of routes, making them never expire

nocturne stag
dusk phoenix
#

Yeah I am aware. Looking at it next week

#

@nocturne stag ideally we would just backport the fix if we know it, to avoid adding new issues in 9.x. For 10.x we'll move to Buildroot 2023.2 with the newer NM

nocturne stag
#

Yeah, I know that "it was fixed some time between VERY_OLD and VERY_NEW" isn't super useful

#

I think if I look at any more NetworkManager code today though I'll have a stroke

#

Hoping one of the NM devs chimes in on their issue tracker

unkempt galleon
#

Hi guys, I was trying to enable hardware acceleration on jellyfin in an addon, and when I checked for amdgpu driver it says it is not available. (I am on Home Assistant OS and CPU is AMD Ryzen 5 5600H, and I am running commands using ssh root on port 22222):

modprobe amdgpu
modprobe: FATAL: Module amdgpu not found in directory /lib/modules/5.15.90

Is this normal / wanted ? Did I miss something ? Would it be possible to integrate it maybe ?
Thanks in advance !

#

Is it only set as a loadable module on x86_64 ? :

operating-system git:(dev)$ grep -nri "CONFIG_DRM_AMDGPU"                                   

buildroot-external/board/hardkernel/kernel-amlogic.config:4292:# CONFIG_DRM_AMDGPU is not set
buildroot-external/board/pc/generic-x86-64/kernel.config:48:CONFIG_DRM_AMDGPU=m
rare sable
#

you need set the option for the add-on for loading modules

#

then you get the modules mapped from the OS

unkempt galleon
#

Ok i didn't know about that feature, i will check it out thanks !

unkempt galleon
#

Mmm the driver I want (amdgpu) is not present on host, so i guess i can't map it to the container ?

find /lib/modules/ | grep -i gpu

/lib/modules/5.15.90/kernel/drivers/gpu
/lib/modules/5.15.90/kernel/drivers/gpu/drm
/lib/modules/5.15.90/kernel/drivers/gpu/drm/i915
/lib/modules/5.15.90/kernel/drivers/gpu/drm/i915/i915.ko
/lib/modules/5.15.90/kernel/drivers/gpu/drm/ttm
/lib/modules/5.15.90/kernel/drivers/gpu/drm/ttm/ttm.ko
#

Should it be added first ? How can I do so ?

dusk phoenix
unkempt galleon
#

Ok thks, any idea when 10.0 will be released ?

dusk phoenix
#

When it's finished 😅 I hope this month, we'll see.

maiden turtle
#

Is there a feature tracking available to see if a certain platform is being supportet (soon) or not?

#

in my case i'd like to run HAOS on an ODROID h3. Last time i've looked there were firmware files missing. Idk if there will ever be a specific version for the Odroid board or if the generic X86-64 will support the hardware

dusk phoenix
#

What firmware was missing last time?

maiden turtle
split sunBOT
#

@maiden turtle I converted your message into a file since it's above 15 lines :+1:

📎 message.txt
tranquil beacon
#

Hi guys, is there any way to get read access to installed addon files? I think the SSH addon puts you inside the container so you can't see those files, right?

dusk phoenix
tranquil beacon
#

Thanks, I was about to try that but found a better way (I think) by using the ssh addon and executing bash into the supervisor container...

blissful nova
#

has any further research been conducted into the issues plaguing netgear nighthawk routers? I understand it's likely the router at fault but I'd like to know what is the actual conflict before i tell them to fix their crap

somber lotus
buoyant token
#

I'm down a rabbit hole... again. I want to create a backup SD Card that I can, in the event of failure of the primary, boot, copy latest backup file from my PC, and restore to be back up and running 100% (with Zwave/Zigbee intact, without having to rename 900 entities). I'm wondering if the following is possible:
(1) Flash new card with latest HAOS using Balena (now I have a bootable card?)
(2) attach primary HAOS card and new card to PC, and using WSL2, cp everything from the primary to the new card.

#

Will this create what I need?

#

And will it boot?

#

I can try it, but need to be sure that putting the NEW card in my Pi won't "reset" anything on the HW that makes it not work as it did when I put the primary card back in and boot.

buoyant token
# blissful nova has any further research been conducted into the issues plaguing netgear nightha...

I have a NIghthawk. What issues are you referring to? I know that at least twice now, my Pi has gone into a state of flooding my network so that I could not access my Pi and overall network traffic slowed. Figured that was fault of Pi, but who knows. I'm also looking to potentially replace my Nighthawk with something with more management capabilities (like subnets), but have always used Netgear.

inner widget
#

I don't suppose OS 10 includes changes to network port management (same supervisor)? Unless something on the host is running on port 1884 now. OS update is the only delta but seem to encounter an issue with the mosquito add-on now because of the port.
Edit: does indeed seem related to the OS update, I see people reporting the issue on dev and rc1 version as well now: https://github.com/home-assistant/addons/issues/2904
Edit 2: Can confirm it is the OS update, rolled back to 9.5 and all is working.

blissful nova
tame musk
velvet sierra
#

HAOS 10RC1 has rendered my mqtt useless. Can't bind to WebSocket 1884. Any solutions as my smart home is now dumb....

#

How do I roll back to 9.5?

blissful flame
#

in the meantime, it's given me an excuse to sqitch to EMQX for mqtt (had to check if it was mosquitto issue or not). Seems both mqtt servers fail in OS 10rc1

velvet sierra
#

Any idea if I can roll back or do I have to reimage and lose 12 hours of data?

dusk phoenix
#

You can roll-back, use ha os upgrade --version 9.5

haughty path
#

Hello, I'm attempting to build haos (dev) for an orange pi 4 lts, I'm getting stuck when it's attempting to build the wifi+bt module (uwe5622). I'm getting this error during output which I think can be solved by appending CONFIG_CRYPTO_LIB_SHA256=y to the kernel .config, however when I add the value to device-support.config, it doesn't add the value to to the kernel .config. However, I do see in the kernel .config file the following: CONFIG_CRYPTO_LIB_SHA256=m but that value is not listed in any of the buildroot-external .config files. I suspect it's being added by BR2_LINUX_KERNEL_NEEDS_HOST_OPENSSL=y in the board defconfig. error log here: https://pastebin.com/Up89UTue

#

any suggestions greatly appreciated

raven cloud
# blissful flame Looks like a known bug (see earlier in Discord). to revert: ha os upgrade --ver...

Version 9.5 is the one listed as last, but when brand new flashed supervisor is not completing the installation. Apparently there’s a 9.6 version that fixes this but it is not available for download.. I would like to setup the backup from my server to do some maintenance on it.. apparently.. the OS version is not stable and it means I’m going probably to install it from scratch (raspberry pi OS / Docker / Ha) .. in general a complete day of work when is just a couple of hours it could be up and running..

copper locust
#

I'm curious about the changes to swap in the new HAOS. I've had some issues of running out of swap in the past and my system locking up. It sort of sounds like we're moving swap to RAM? But I thought swap was basically just extra memory on disk, so it kinda sounds redundant. But I'm sure I'm missing something so would like to get informed 🙂

dusk phoenix
#

So OS 9 used zram which essentially compresses some memory (technically, it uses swap mechanism for that: the pages the OS decides to swap out are actually just compressed and still stored in RAM).

With OS 10 we use zswap, which is just like regular swap on disk, but compressed (to maximize the amount of memory which can be stored, while preserving write cycles of the storage)

copper locust
#

Ah I see. So OS 9 actually did do away with on disk swap?

dusk phoenix
#

Even before 9, Home Assistant OS didn't use swap on disk since pretty much the beginning

copper locust
#

Wait so OS 10 is going to be the first to use on disk swap?

#

Because I'm almost certain I've run into memory leak issues (unrelated) which spilled over into swap, and then when my swap filled up my entire system locked up and sometimes I had to hard reboot.

#

I'd actually be curious on your thoughts, but I was considering disabling swap altogether on HAOS. A lot of people seemed to suggest against it, but it leads to serious consequences when there's a memory leak and my swap fills up. With all the disk activity it completely locks up the entire OS sand I often have to hard reset it which can be dangerous.

I could increase my swap size, but unless I'm monitoring swap usage a memory leak could still fill it up. The alternative is turning swap off and processes just run out of memory and get killed. This alternative seems better than the entire OS locking up.

What are your thoughts on this?

dusk phoenix
#

So with that, the system will kill processes which use too much memory (most likely the app which leaks memory) quickly.

copper locust
#

But I never confirmed it was actually IO related.

#

Is there a way to see IO usage per process? I've used iotop before, but that isn't available on HAOS. Would be cool maybe to even have a sensor on Supervisor similar to the cpu and mem ones

dusk phoenix
#

You can docker exec -it homeassistant /bin/bash and then apk install iotop 😎

copper locust
#

I could yeah haha

distant sorrel
#

Hi, where should I look to find more about the OS upgrade process? Like how's it knowing whether there's an upgrade or not. If yes, how's the upgrade done.

raw spire
distant sorrel
copper locust
#

Are core addons tested as part of the HAOS pipeline?

dusk phoenix
#

Are core addons tested as part of the HAOS pipeline?
Not really. Some add-on also need accounts/configuration to make it an useful test.

tepid hollow
stoic bison
#

@tepid hollow Yes, I took a screen shot of the transition from OS9.5 to 10. CPU utilization per Proxmox doubled. 5% steady state to 10%.

tepid hollow
#

Yeah see first imgur above. I had the same experience

#

From about 2.5% to about 5.5%

stoic bison
#

Mine went from 5% to 10% right after the reboot to OS10 and has stayed there 24 hours later.

copper locust
#

This is CPU usage of the host or core?

tepid hollow
#

We're both virtualised, so it's the VM CPU usage, which seems to be (at least for me) the ha container

stoic bison
#

Yes CPU utilization of the HAOS VM on Proxmox

tepid hollow
#

according to top (showing containerd) and "docker stats" showing mostly homeassistant.

copper locust
#

Those are two conflicting answers

#

I assumed it was the host VM as @stoic bison said. Is it that or the container for core?

tepid hollow
#

There's multiple "hosts" here 🙂 They all bubble up. High CPU in homeassistant container causes high(er) cpu usage on HAOS (VM) causes higher on host (proxmox)

#

See my imgur posts for my quick analysis.

stoic bison
#

I'm running HAOS as a real VM, NOT a LXC container or Docker

tepid hollow
#

Same. But HAOS runs containerd.

stoic bison
#

I used tteck's HAOS VM deployment script.

tepid hollow
#

Same

stoic bison
#

Yes within HAOS containers are running

tepid hollow
#

But "problem" is at homeassistant container, everything "up" from that is symptom.

#

See my screenshots man 🙂

stoic bison
#

ok I didn't delve into the container level...I'm just looking at HAOS VM level stats

copper locust
#

Well the screenshots don't show a comparison. The only comparison is the graph and that graph doesn't say if it's the host or container CPU usage

tepid hollow
#

I'll dig up the container stats, sec.

copper locust
#

The screenshots showing processes are just pictures of %'s

stoic bison
#

My screenshot is at the VM level (HAOS) per Proxmox. it has no knowledge of containers running inside the HAOS VM.

#

It's the hypervisor view of the VM CPU utilization...my screenshot was NOT the Proxmox host level it's VM level graph

#

My Proxmox host has 12 physical cores, 16 threads (12th gen). So HOST level CPU is tiny (2%), but the HAOS VM doubled it's CPU resource usage with OS10

#

HAOS VM is assigned 2 vCPUs

tepid hollow
#

fwiw The highest CPU process on HAOS login appears to be the python3 -m homeassistant --config /config process within the container.
(though supervisord is using a bit too)

copper locust
#

I don't use Proxmox, but thought I'd look into it anyways. It actually looks like since the update my CPU usage is at the lowest it's been in at least a month

#

Though it is climbing

#

I just updated last night at 11pm though

stoic bison
#

My CPU doubled and stayed doubled right after the reboot into OS 10

tepid hollow
#

Ditto

copper locust
#

I run in a VM as well, but I just use straight vbox

stoic bison
#

@tepid hollow What command are you running to show that top looking stats?

tepid hollow
#

From haos console type "login" to get linux prompt, then "top" to run top; or "docker stats" to get info about containers.

stoic bison
#

ah I have protection mode on......guess I can turn that off

tepid hollow
#

protection mode? use the proxmox console, not ssh

latent wigeon
#

Either is fine for this

stoic bison
#

I'm' confused...Proxmox host won't have any info on Docker containers inside HOAS VM.

latent wigeon
#

The console for the VM

tepid hollow
#

you're not in proxmox by then. just using the "console" of the VM to get into the VM console, from there using docker tools.

stoic bison
#

Yes I'm SSH'd into HAOS

tepid hollow
#

not the pve node console, the vm console

stoic bison
#

yes....won't let me run docker stats due to protection mode being on

#

I have to disable that within HA UI first

tepid hollow
#

Hence not using ssh

stoic bison
#

well I'm not going to mess with docker.

tepid hollow
#

You don't need to mess with it. "docker stats" and "docker logs" is harmless.

tepid hollow
#

So... any tips on how we can help? This higher CPU usage seems unexpected? (and showing up in container, doesn't point to impact of swappiness, right?)

tepid hollow
tepid hollow
radiant jasper
#

I had to retirned to 9.5 .... with 10.0, my CPU jumped from 6% to 30% (Hassio/Odroid C4)

tepid hollow
#

Yeah I reverted too

stoic bison
#

I'm still on 10, but ya something is sucking up CPU and needs to be debugged.

leaden epoch
#

Was HAOS 10.0 removed? I did not update and no longer see it available in the update screen

tepid hollow
#

It was pulled from some platforms, as it could brick boxes, or fail the upgrade

#

FYI (and maybe this is worth a pin), if you're having problems with HAOS 10.0 and your box is not bricked, you can revert to 9.5 with ha os update --version 9.5 (skip the "ha" if you're at the homeassistant prompt, not a shell prompt)

tepid hollow
#

So... devs, how can we help with this issue?

tepid hollow
#

I might profile before vs after, see if anything sticks out. Unless y'all have a better idea?

dusk phoenix
#

The general CPU higher usage due to containerd is tracked by: https://github.com/home-assistant/operating-system/issues/2476

I haven't really started investigation, but it seems that this is just caused by the update to Docker 23. I haven't found a related issue in the Docker issue tracker. We'll have to figure out if this is a generic Docker 23 issue or something which is specific to our build.

zealous wing
radiant jasper
#

Temperature increased about +3-4 degrees

blissful nova
tepid hollow
#

The "increased CPU on 10.0" issue doesn't seem to be getting much traction. How can we help? (as competent sysadmin/python devs, but not especially familiar with HAOS)

rare sable
#

You can open an issue on containerd and ask how to tackle it and if they interested to solve the 2-3% more usage

zenith comet
#

I figure it's related to how it's used within HAOS though, otherwise there would have been others complaining

tepid hollow
#

And it's not 2-3% more usage (of host). It's double.

#

which is not exactly trivial (power, heat, etc)

tepid hollow
#

There's 5000+ commits between docker 20.10.24 (HAOS 9.5) and 23.0.3 (HAOS 10.0). Too many to review, and I don't see anything obvious looking in the docker/moby issue queue...

wintry depot
#

Hi. Having an issue with bluetoothd. All my devices and integration that relies on bluetooth keeps dying after a couple of mins. I need to 'refresh' the bluetooth integration for it to work and goes dead again. I'm running HAOS on a NUC and close inspection of the host logs I can see lots of errors May 03 08:08:24 homeassistant bluetoothd[438]: Unable to create object for found device 00:00:00:00:00:00
May 03 08:08:28 homeassistant bluetoothd[438]: src/device.c:device_new() Unable to register device interface for 00:00:00:00:00:00

stoic bison
#

@tepid hollow My HOAS CPU more than doubled from ~3% to 10% with the 10.x update. That's not trivial.

tepid hollow
#

I agree

onyx prairie
dusk phoenix
#

I did not. I'll have a look at it next week.

onyx prairie
misty plume
#

Hi all, I was wondering since I can't find this issue in the bug tracker.

I run Home Assistant in Proxmox via a Intel Nuc. On version 2023.4.4, all works good. But, after upgrading to a higher 2023.4.* version (don't know which) some (not all!) Mi Flowers are not working anymore. Some flowers go down and up. Also does the HA SkyConnect not work propertly. Some devices are not reachable at all, and after some hours, all Zigbee devices go down. Only a stop/start in Proxmox for the home assistant guest will work.

Tried to update to versions 2023.5.1 does not fix this issue. Reverting back from a Proxmox backup to 2023.4.4, all is working as expected again.

The pity part, I don't get any specific logs. Is this a known issue?

raven cloud
#

Great Job Guys On PI4.. Numpy and OpenCV can not be install.. it fails to install the Numpy part apparently.. using SSH addon is there the possibility to enable docker commands?

zenith comet
raven cloud
zenith cypress
#

not sure if this is the right channel, but I was curious what the lag time is before HomeAssistant OS installs tend to move up to the newer python version once core supports it? I noticed that 2023.3 supported Python 3.11, and I attempted to use an enum.StrEnum in my new integration, which promptly failed miserably once I tested it on my HomeAssistant OS based install, because it appears to still be using python 3.10.11

#

not a complaint at all, I was just curious, I tried digging through several repos, but I'm not sure which one is responsible for building the container images that are run in HAOS

tepid hollow
#

@zenith cypress looks like dev image already using 3.11.3

zenith cypress
#

oh, nice, so it's literally right around the corner!

rare sable
#

as we have over 1000 library that we depending, it takes 1y until all important libs supporting a new python version and all bugs are also fixed in python (mostly starting with .4)

#

once we feel it's stable enough, we update

mossy seal
#

is there a way to add raspberry pi config settings to the uboot/alpine distro? my little tft is upside down and i need to add something to invert the screen

haughty path
#

@mossy seal are you asking about alpine linux or home assistant operating system?

zenith cypress
# rare sable as we have over 1000 library that we depending, it takes 1y until all important ...

Yep, I understand that, I wasn't asking about the general delay between when a new python version comes out and when Home Assistant will support it, I get that will take quite some time as all of the underlying libraries need to be updated as well. I was more curious about the time between when Home Assistant supports it, and when HomeAssistant OS will ship using that new version. Looks like it was about 3 months in this case.

stone mason
#

Bluetooth: hci0: Failed to load Intel firmware file intel/ibt-0040-1050.sfi (-2) on Beelink S12 Pro OS 10.1.

stone mason
#

Same on 10.0 and 9.5. Is it worth going back farther?

haughty path
#

Intel claim the AX101 supports linux, but they don't provide a driver for it

mossy seal
haughty path
coarse silo
#

Hey folks. Not sure if this already has been asked/answered, but ever since the last HA-OS update (from 9.5 to 10.1) I seem to have startup issues on full reboots. I run HA-OS on a RPI4 connected to a multi-bay USB3 HDD enclosure (externally powered) and have moved the data partition to an SSD inside this enclosure. In there, there are additional HDDs (I had spare) which I use for media (pseudo NAS). On reboots I have to unplug all HDDs other than the SSD, or HA simply won't boot. It appears as if it tries to check what's on those discs, crashes and is starting over again. I have no screen attached, so I don't see what might be printed on the console, but I can see f.e. one of the HDDs spins up, has read access for a couple secs, then all HDDs seem to power down and immediately up again (like on a reboot) and the loop continues endlessly

coarse silo
#

I am not booting from USB, it's still booting from sdcard, only the data partition is moved to the USB SSD

dusk phoenix
#

We've updated U-Boot, so maybe U-Boot hangs. Connecting a screen would be interesting to see what happens/shows up. A issue with all the information collected would be ideal 😄

coarse silo
#

@dusk phoenix can it still be a USB SDD boot issue when my boot drive is still the sdcard and only the data partition has been moved to the SSD?

#

can try to get a display attached (I only have old DVI screens in the server room, so need to search/find a DVI->HDMI dongle)

dusk phoenix
hearty parcel
#

Hi Folks, I just upgraded to OS10.2 and have noticed that CloudFlared remote access is failing. This is appearing in the logs which may be related '''Logger: homeassistant.components.http.forwarded
Source: components/http/forwarded.py:125
Integration: HTTP (documentation, issues)
First occurred: 11:13:32 (715 occurrences)
Last logged: 11:40:40

Received X-Forwarded-For header from an untrusted proxy 172.30.33.3''' While I'm figuring it out is there a recommended way of rolling back the OS version?

boreal elk
#

Hi all,
I've got this Asus mini pc PN41, with Home Assistant running on it from boot.
It's been running fine, but yesterday system became unresponsive and I saw a bunch of: EXT4-fs error. Similar to these in this thread for Ubuntu: https://askubuntu.com/questions/1100838/ext4-fs-error-device-sda2-ext4-find-entry1436-ubuntu-18-04

The SSD is an nvme from Kingston. Temperature never got above 60 as far as I can see. Processor use not above 30%. Memory at 16%.

I did an fsck afterwards via a Mint live USB and that was all OK. Check in bios on nvme also passed.

Thus far I've not yet been able to reproduce this issue.

Any experiences with this?

plush dagger
#

Hey. I'm investigating slow boot on Odroid M1 (w/ RTC battery) and as first slowdown I can see:
[FAILED] Failed to start Wait Until Kernel Time Synchronized.
[FAILED] Failed to start D-Bus System Message Bus.
[FAILED] Failed to start Home Assistant OS Agent.
[FAILED] Failed to start RPC Bind.

boreal elk
#

I've seen a failed to start RPC bind also. But was not slowing down startup.

zenith comet
#

If you post this in the right channel you might get some help diagnosing it 😉

plush dagger
tepid hollow
zenith comet
tepid hollow
#

Thanks, just read that. Doesn't see to be going anywhere. It's "strange" that updating docker caused CPU load to roughly double for lots of people. Impacts power/heat/sociability of HA(OS)...

rapid compass
#

Hi, is it possible to add something like picture element to homeassistant but for wearos system? To create such tiles, the current tiles and shortcuts are poorly developed and functional. If you could create the possibility of adding 5 tiles and on each tile it was possible to create something like this, that would be great.

onyx prairie
tepid hollow
#

Hmmm, it takes a really long time to build HAOS, eh?

tepid hollow
#

I'm reading the buildroot docs; but is there a TLDR for dummies who just want to configure/build a custom haos?

zenith comet
tepid hollow
#

Ya, I'm following/building that now. any way to estimate progress? It's up to host-gcc-initial-11.3.0 in ./output/build

#

2.25 hours so far

zenith comet
#

If that were possible it would probably be enabled by default. Cmake could sort of do this afaik

tepid hollow
#

I guess if I could figure out all of the components it was building, at least I'd know I was 24-of-something 🙂

#

Installing an ubuntu-server VM elsewhere, that I can give more resources, rather than my poor old mac, which is sharing with all my real-work stuff

tepid hollow
#

OK got my custom images built. Way faster on linux than on mac.

#

25G disk requirement tho! Ouch.

dusk phoenix
#

Building from scratch takes typically takes 30 minutes or so on my machine. But yeah, it requires disk space... All the object files 😰

tepid hollow
#

The first time is apparently quite a bit slower. Now I need to figure out how to change things 🙂

dusk phoenix
#

There might be space saving options in Buildroot, but it comes at the expense of slower rebuilds.

tepid hollow
#

It's no big deal, i just had to resize my VM to make room

tepid hollow
#

Hmm, all my CPU% are reporting lower than they were yesterday. No idea why! Didn't reboot/upgrade proxmox, and 3/4 of the VMs are the same ones I was using yesterday. 🧩

shell junco
#

hmmmm - how can I run a trim on the host os?

zenith comet
#

Use the fstrim command and a proper channel

haughty path
coarse silo
# dusk phoenix We've updated U-Boot, so maybe U-Boot hangs. Connecting a screen would be intere...

sorry that it took a while, but I finally got a friggin micro HDMI adapter and was able to connect a screen to my RPI4. Unfortunately the screen stayed blank during the reboot loop. I didn't even get to see the usual RPI4 startup rainbow screen or alike. Only after I have removed all HDDs other than the OS SDD from the multi-bay USB case and the startup process continued, it did print things on screen. Not sure what else I could do to debug this issue 😦

rare sable
#

It will properly work more stable if you run the OS from SD card and offload the data partition (we call this feature data-disk) to the SSD over USB

coarse silo
#

that's how it is setup

#

only the data-disk is on the SSD. OS remained on sd-card

rare sable
#

all HDDs other than the OS SDD you mean the OS SD?

coarse silo
#

ok, I just power cycled the RPI again and had the USB case powered off (which contains the data-disk SSD as well as two HDDs for media). Then the system did boot from the sdcard as it should and waited on the drive with the data partition to get online. After I turned on the USB case, startup continued just fine. So the issue seems to occur way before the OS tries to boot, so likely uboot then?

rare sable
#

no output is strange, seems the boodloader hung because of the USB initialize

coarse silo
rare sable
#

no output means, RPI not load our OS - so it stock in a very early state

#

however, our OS also right now only support 1 external disk, the datadisk

coarse silo
#

the boot loop cycle remains until only the data SSD is still in the USB case. As soon as any other HDD is in there, I can see (and hear) that the HDD is spinning up, has random read access and then power cycles (likely due to the system rebooting).

rare sable
#

RPi have this USB boot feature that force them to understand and load USB during early boot. It seem that hang now

coarse silo
#

well, it worked just fine with all HDDs until the recent bootloader changes IIRC

rare sable
#

but is known, that some USB disk extender make issue or work and after we update the RPi firmware, it stop working

#

we not have our own RPi firmware, we use the official firmware, but if you google, that USB boot stuff cause a lot of issues, even if you not like to boot from USB directly

coarse silo
#

if it is a RPI firmware thing I should be able to disable something in the config.txt, right? Maybe turn off USB boot?

rare sable
#

maybe, you need lock at the RPI documentation

coarse silo
#

or some dtoverlay parameter. Googling right now

rare sable
#

but if you realy like to build a NAS or disk like system, properly with an board that have real NVME or SATA support, you will be more happy over the time as with an DIY project board

#

you can also but the old firmware on the boot partition hand hope RPI fix the bug in the firmware (we are quite up-to-date to sync that)

coarse silo
#

I have a real NAS, the PI only should contain temporary stuff that is available 24/7 without the need for failsafety (so no ZFS, RAID, etc)

#

so basically giving some spare HDDs a second life/use

dusk phoenix
#

Best thing you can do to debug is enable/uncomment enable_uart=1 in config.txt of the first partition of the SD card

#

that will give you a serial console on GPIO 14/15

#

There you should see if the early firmware has issues with your USB disk or U-Boot.

coarse silo
#

I don't have the gear to connect a serial interface to GPIO 14/15

coarse silo
#

if it is a firmware bug, the same issue should show up when I try to boot raspberryPI OS?

plush dagger
#

hm… Every start same problem. (recursive deps?)
dbus-broker.service

01:04:04 homeassistant systemd[1]: Starting D-Bus System Message Bus...
01:05:34 homeassistant systemd[1]: dbus-broker.service: start operation timed out. Terminating.
01:06:08 homeassistant dbus-broker-launch[270]: Ready
01:06:08 homeassistant dbus-broker[497]: Dispatched 5 messages @ 31(±23)μs / message.
01:06:08 homeassistant systemd[1]: dbus-broker.service: Failed with result 'timeout'.
01:06:08 homeassistant systemd[1]: Failed to start D-Bus System Message Bus.
01:06:08 homeassistant systemd[1]: Starting D-Bus System Message Bus...
01:06:08 homeassistant systemd[1]: Started D-Bus System Message Bus.
01:06:08 homeassistant dbus-broker-launch[502]: Ready

Full info: https://dpaste.org/xdiJ2

split sunBOT
#

@plush dagger I converted your message into a file since it's above 15 lines :+1:

📎 message.txt
torpid wigeon
#

Using fsync inside an add-on will really write file to SD-card? I know that ext4 commit interval is 30s. I would like to save new data and then delete old, but I can do this only if the new data is really written to storage (eg. SD card). Can it be achieved with fsync inside an add-on's container?

zenith comet
#

That interval is not for fsynced things afaik

plush dagger
#

upd: nvm. I'll just wait until someone fixes it 🙂
Hi. Is it possible to use "release SD card" and change some services on it?
I'm trying to locate bugs on Odroid M1

quick pulsar
#

Hi Guys, I have a question relating harware but also OS i think, im trying to get some bios information, and see that you can use dmidecode to see if its avaiable, so ultimately i am trying to get it from node red, and am after the serial number of the device if possible, but dmidecode is not installed natively, is there another way to achieve this, i have tried to search the OS git, but there is just one referece saying that it wouldbe cool to include, but nothing furhter, seaching serial brings a bunch of unrelated issues :p anyone with any direction that i can look at?
I am running the Generic X86-64 build, not sure if this is the right forum?

tepid hollow
#

Do you need to know something about your particular system, as a one off? Or have some repeated need? @quick pulsar

quick pulsar
#

Hey @tepid hollow if i can do it once i can do it many times ;p, but yea, its helpful for me to know to match the serial number to the instance, but i am sure there is many real world examples for others barring knowing what unit it is when you have multiple same hardware. But i am considering it for comercial a comercial application, and it can help a ton if someone just provides a serial number on the device as apposed to some arbitrary reference that you cannot confirm via software

tepid hollow
#

Time to learn build, or how to make a static binary

quick pulsar
#

me?

#

well i mean there seems to be tools available for it, but everywhere i look it needs to be included in the OS Roll, since the OS is mounted RO

#

Buildroot, the tool used to build the Home Assistant OS, is designed to create a complete, standalone custom Linux system for an embedded system, and this happens at compile/build time. It's not a conventional Linux distribution and doesn't include a package manager like apt, dnf, yum, or pacman for installing packages on a running system.

The operating system it generates is essentially a read-only system with very limited write areas. Because of this, you can't directly install additional packages onto a running Buildroot system in the same way you might with other Linux distributions.

If you want to add a tool or utility such as dmidecode to a Buildroot-based system like Home Assistant OS, you would generally need to add it to the Buildroot configuration, rebuild the system image, and then flash that new image onto your device.

#

Run make menuconfig to open the configuration menu.

In the menu, go to "Target packages" -> "Hardware handling" and select dmidecode. This can typically be done by using the arrow keys to navigate to the option and pressing the Space bar to select it.

Exit the menu and save your configuration when prompted.

Run make to build your system.

#

So thats how i ended up here, since i assume you guys already have the toolchain, i really dont want to get version locker, the other option for me would be to roll a add on, but i dont know how effective a docker will be at getting low level information of a system, and then it becomes an unsafe add on, but potencially possible to do it this way

#

something like this:
FROM debian:latest
RUN apt-get update && apt-get install -y dmidecode
CMD ["dmidecode"]

#

and then just push that via mq

#

or i guess i could push it direct to HA with the API at that point as well

tepid hollow
#

I played with buildroot but didn’t learn enough to make real changes.

quick pulsar
#

apparantly you can just run
dmidecode -s system-serial-number

#

but yea, i need to confirm that as well,

#

ok well interesting place i find myself

tepid hollow
#

Have fun! 🙂

#

I was exploring a docker/kernel performance issue with os 10.0

#

But not getting much help from “real devs”

rare sable
quick pulsar
#

ok thanks, that gives me some direction

shell junco
#

So I have no idea where to post this - as there isn't really a HassOS room, and a search of the discourse site has turned up nothing....

I've set up the new network storage to an NFS storage - but every time the entire VM boots, the NFS volumes always fail to mount. If I hit the "Reload" button on each storage, it maps it again and it then works fine until the next VM boot.

Is there an issue known about this? Or where I would start to look at the underlying problem (given I can't figure out how to get into the host)

sinful heron
shell junco
#

yeah - same deal - I think HA starts and expects it to be functioning before the host OS maps the volume.

#

just had a mini earthquake happen too lol

tepid hollow
#

Melbourne?

shell junco
#

yep

errant heron
#

hello everyone, I encountered an error while using sudo scripts/enter.sh make odroid_m1 to compile the os. Could someone please help me with resolving it?

# 
Writing superblocks and filesystem accounting information: done 

Waiting for Docker daemon...
Loading container images...
Loaded image: ghcr.io/home-assistant/aarch64-hassio-supervisor:2023.06.2
Loaded image: ghcr.io/home-assistant/aarch64-hassio-audio:2023.04.2
Loaded image: ghcr.io/home-assistant/aarch64-hassio-cli:2023.05.0
open /var/lib/docker/tmp/docker-import-2622527219/repositories: no such file or directory
make[1]: *** [package/pkg-generic.mk:374: /b
haughty path
#

@errant heron did you try again? GitHub was down for a while, could be that

split sunBOT
#

@ocean tide I converted your message into a file since it's above 15 lines :+1:

📎 message.txt
ocean tide
#

Whether there is documentation for compiling os images

zenith comet
rapid compass
#

Are proposals for applications for the WearOS operating system still considered in such a way that there is a chance that they will be accepted and implemented?

strong ridge
#

"Add Network Storage"

#

Is anyone else having issues adding a folder called "frigate"? I keep getting the error "Cannot mount bind_frigate at /data/media/frigate because it is not empty"

#

I have tried adding and readdding many different ways, and my other 3 network storages work fine (Media, Share, Backup)

#

Also the "frigate" folder on my NAS is brand new, as in I just created it... hence it is empty

strong ridge
#

*** fixed it ***

#

It seems I need to SSH into HAOS, then manually find a local directory that was created by Frigate and delete it. Then I can add this

stray scarab
#

Hi Folks, sorry in advance if this isn't the right place to ask for help with an issue but I'm currently stuck unable to upgrade my HAOS to 10.x and not sure how to fix this. I've raised an issue on github but its gone a bit stale and I need some help https://github.com/home-assistant/operating-system/issues/2522

tame musk
split sunBOT
#

@fervent ravine I converted your message into a file since it's above 15 lines :+1:

📎 message.txt
fervent ravine
#

So this morning i was picking up my new coral usb device, i plugged it in my raspberry 4 8gb, and configured it in frigate, after a short period of time my ha interface was crashing, not loading and not working. I tried connecting to terminal using ssh but no success there either, so i pluged the raspberry pi in and out so it would restart, but there was no home assistant starting after this, instead it shows the raspberry pi bootloader screen with some error. Anyone know how to fix this ?. Bootloader screen info:

split sunBOT
#

@fervent ravine I converted your message into a file since it's above 15 lines :+1:

📎 message.txt
haughty path
rapid compass
#

Hi, is it possible to somehow display the image from the camera on the galaxy watch 4 lte?

latent wigeon
#

Are you developing that?

rapid compass
#

Yeah, sorry

somber island
#

eta on dns fix?

dusk phoenix
#

Context?

somber island
#

github open issues with dns

dusk phoenix
#

We have 102 repositories and several thousands of open issues 😅

faint stream
eager yoke
#

I am trying to build HAOS for a new board, Sonoff iHost with Rockchip rv1126

Right now the u-boot-dtb.img is not landing in the correct location. I need to manually dd this to get it too load. What controls the location of this?

eager yoke
#

never mind, found that

dusk phoenix
#

Sorry for the delay, some 🦠 took me out the last few days 😢 I'll respond to the issue.

faint stream
eager yoke
#

Is it possible to override the OTA url for updates from within the HAOS rootfs ?

#

I have the Sonoff iHost running HAOS now btw 😉

haughty path
#

@eager yoke let me know when you figure it out 🙂

eager yoke
haughty path
#

Updates

eager yoke
#

ok

haughty path
#

Thanks

dusk phoenix
#

The OTA URL is part of the version file (see https://version.home-assistant.io/stable.json). The location of this in turn is encoded in the Supervisor. In general, the updates of all parts of the system (including the OS) are coordinated by the Supervisor.

eager yoke
forest horizon
#

I've migrated my Pi4 instance twice now to virtual servers, first on TrueNAS (which I found my box just wasn't quite up to spec to handle) and now to Unraid. With both I encountered similar issues of the restoration process freezing up and needing a few pokes through the CLI over the VM's VNC server. And in both cases I had add-ons fail because they needed to be rebuilt for the new architecture (at least, that's why I assume the rebuild was needed).

I feel like the above, whilst frustrating, probably wouldn't be called a bug but could be informative for the onboarding + restoration flow development. Where would be the most sensible place to open a ticket to describe in more detail my experience and is there a suggested template for such a thing as I'd rather not clog up with a bug report if it's not the best platform? 🙂

worthy prawn
#

I have problem with Backup, I don't see it backup in the setting page and when create backup not change name file that I entered (both manuali backup and with automation backup)

inner widget
worthy prawn
dusk phoenix
maiden turtle
#

question: If I remember correctly there was a way to add missing modules for the kernel via an USB stick to HAOS. Is this still existing?

eager yoke
dim rover
#

Possibility of setting cpu governor on odroid M1

velvet sierra
#

Where in HAOS, either within the system image (I'm running on VirtualBox) or in the build repository (https://github.com/home-assistant/operating-system) is specified which images launch upon boot? For example the image qemux86-64-homeassistant:2023.10.3 . I must be missing something very obvious.

#

Suppose I want to patch that image with some altered Python files, is there a handy /established way to do that that is faster than rebuilding the image? Just a new Docker layer?

dusk phoenix
#

This is controlled by the Supervisor

#

There are Supervisor command to use a different image: ha core options --image <your image>, then ha core rebuild, but its a bit a pain to use as you need to have the same version as your original image under that name.

#

I recommend using a Visual Studio Code Dev container foor development. If you want to deploy something custom in Core onto your OS installation, use custom_components.

velvet sierra
#

I recommend using a Visual Studio Code

mossy seal
#

having a problem upgrading core to 2023.10.5 on my rpi3 haos install. i think i've narrowed it down to the docker pull attempt crashing at the 573580e7c1c0 layer with:

failed to register layer: error creating overlay mount to /mnt/data/docker/overlay2/70f11ab31addf342a4ee7efbdfdae93a85b6bff80e90c43d8c4168c5c1772a40/merged: too many levels of symbolic links

#

anyone else report this? i'm on 2023.10.3 now and this has only just started happening

zenith comet
wide rain
#

Out curiosity, why are docker processes not showing on the host when running ps or top? I have looked for other process namespaces, but found none other than the host namespace. I'm running HA OS.

zenith comet
#

Where, or better, how are you running this? Remember that the SSH addon is a container.
This is likely not the right channel, btw

wide rain
rare sable
#

ha su repair should fix it

worldly crystal
#

I have a general question about HAOS development - I've been battling stability issues on recent HAOS versions (see https://github.com/home-assistant/operating-system/issues/2782) and am coming to the conclusion that it's related to recent kernel updates related to the i915 chipset. For now, I've had to rollback from 11.1 back to 11.0, which seems pretty stable for me.

My question is a more general one - HAOS seems to always run the absolute, latest version of the Linux Kernel (6.1.61 has just been pushed, which was only released a week ago), and while this is a noble goal, in any other walk of life this sort of bleeding-edge software would not be considered for critical infrastructure as, at best, it would be considered unproven. I guess the discussion topic here is whether this is actually a good thing to do, particularly if it's possible that it'll introduce bugs that are entirely unrelated to HA. Is the product being unnecessarily risky, especially given that for many, stability is far more important than having the latest, cutting-edge features?

dusk phoenix
#

@worldly crystal we track the latest LTS (long-term support) Linux kernel (which is not the absolute newest kernel, that would be 6.6.1 at this point). We are not really after the newest cutting-edge features here (and also don't get them, since this are stable kernel updates with bugfixes and security fixes). From upstream Linux, there is no "better"/more stable kernel to track (at least not that I am aware of).

For now, I've had to rollback from 11.1 back to 11.0, which seems pretty stable for me.
That is exactly the reason we have the A/B system, which allows us that easy rollback 🤷‍♂️

I think part of the problem is that people run HAOS quite often on some older/more obscure hardware, where other distributions aren't used that often (or an older version is being used).

Do you have another suggestion what to track (while getting security updates)?

coarse kiln
#

Does the x86-64 kernel contain the uas driver and if not, is there a reason why?

Long story short, my Odroid N2+ h/w started dying and I repurposed an old Intel-based chromebook (similar h/w to a NUC) to run HAOS using my existing USB3 SSD data disk. All working well after flashing UEFI firmware but the SSD data disk is using the usb_storage bulk transport driver rather than uas which I'm fairly sure was being used by the odroid_n2 config. The chromebook ports are all USB3...

[    0.610031] usbcore: registered new interface driver usb-storage
[    2.021599] usb-storage 2-4:1.0: USB Mass Storage device detected
[    2.030022] usb-storage 2-4:1.0: Quirks match for vid 174c pid 55aa: 400000
[    2.038701] scsi host0: usb-storage 2-4:1.0```

The quirk isn't related to UAS compatibility.
worldly crystal
# dusk phoenix <@695356475819819020> we track the latest LTS (long-term support) Linux kernel (...

Thanks - and the quick rollback is very useful and works very well. I've just hit issues that, stable or not, look to be recent kernel introductions (and I'm hoping will improve in future versions). I'll just keep trying each new version and will rollback if not stable! Thanks for the info on the 6.6.x vs 6.1.x.
Regarding my own issue - it's one of those things that I would have expected to be more widespread, but as it's not must clearly be related to my own hardware, but it's hardly obscure or terribly dated (Dell Optiplex i5-8500T), but I cannot clearly explain why it's become unstable in recent time, yet is stable when rolled back a bit. I guess I'll just keep trying new versions and see what happens!

rapid compass
#

Hello, I see that a new option has been added - the camera tile. I know that I have already asked this question about the refresh rate, but is it possible to cheat the system and have the tile refresh every second? An additional thing that pretends to click on the screen.

#

I want as many new features as possible for the wearos system because it is the most convenient thing that can be. I dream of creating a monitoring tile that will allow me to view live monitoring with a slight delay.

dusk phoenix
#

This channel is about OS development. Your question better fits to #frontend-archived probably

steep crane
#

@dusk phoenix where do you want comments on the RCs? By the way no issue for me on rc 11.2

dusk phoenix
stray gazelle
#

Does anyone know if there will be an alpha or beta release? I would be happy to volunteer as a tester.

latent wigeon
feral pewter
#

Hello, I am running yesterdays development build on the rpi5. How can I force install add-ons on dev builds?

sterile python
#

I don't know if this is the right channel, but I was wondering if there was any consideration of making an ISO / installer for the generic x64 version of HAOS?

dusk phoenix
#

Yes we are considering it. We have a buildroot based installer for Yellow and Green (sources here https://github.com/NabuCasa/buildroot-installer), we could make one for x86-64 too. The ones for yellow/green are headless (as in, nothing really shown on the HDMI/serial terminal) and just install to the internal eMMC by default. I guess for x86-64 we'd need some type of user interface so the user can choose the target disk etc. It would help setup EFI variables right (solving issues such as this https://github.com/home-assistant/operating-system/issues/2352).

sterile python
#

Awesome! Thanks!

white quiver
#

Small question to the developers: I installed the 11.3.rc1 on the RPI5. What are the implications when the final version comes out? Do I need to reflash the image or is there an OTA option to go stable?

stone bough
#

There are OTA updates, so no need to reflash

white quiver
inner widget
white quiver
#

Thanks a lot everyone for your help developing HAOS for the RPI5, I have the stable version running and life is good

somber lotus
#

Just done the OS update and now all I am getting is The following integrations and platforms could not be set up:

template.switch

#

Invalid config for 'switch' from integration 'template' at configuration.yaml, line 52: 'sensors' is an invalid option for 'template.switch', check: sensors, please check the docs at https://www.home-assistant.io/integrations/template is full error, was this meant to be coming and I have just forgot as I have frazzle brain still as someone gave me the present of covid for christmas

slim atlas
velvet mantle
wise bolt
lament star
#

~~FWIW, I submitted a PR for an update to grub.cfg - this relates to issues with x86 updating OS's and proper Slot selection on boot of the hardware https://github.com/home-assistant/operating-system/pull/3050~~
I am going to retract this since I am now realizing this requires additional configuration in order to work.

eager yoke
somber lotus
lucid tangle
#

Hi, Have anyone got HA running on a USB SSD M2 on a R Pi5 yet? Mine keeps chrashing after the second reboot. Even with config.txt loaded with usb_max_current=1 . Yes and I have a 66W USB-C supply.

lucid tangle
raven cloud
#

Guys any possibility to use OpenCV on the OS based implementations? the question is related to development of an image processing component.. I didn't try yet if I can import OpenCV with the new OS releases.

#

never-mid I just put in the requirements "opencv-python-headless==4.6.0.66" and it fails to load.. thanks anyway.

rare sable
#

we did support it, but generate wheels for it was very dificult and time intensive

raven cloud
nocturne cypress
#

Any clues on how to get ssh to the OS on a Yellow working? I followed the instructions here: https://developers.home-assistant.io/docs/operating-system/debugging/ to enable ssh access to the OS, but port 22222 remains closed, even after a reboot. I believe I need this access in order to mount a drive, which I can see from the regular ssh prompt, but can't mount.

dusk phoenix
nocturne cypress
#

thanks - can you point me to where I might find that web front end? I see Supervisor in integrations, but no obvious user interface for it.

dusk phoenix
#

It is all integrated as part of the Home Assistant frontend. The mount options are in Settings > System > Storage > Add network storage

nocturne cypress
dusk phoenix
#

Ah yeah we do not have a built-in solution for that 😢

#

Probably closest work around is to use Debian 12 + Supervised.

nocturne cypress
#

A quick look at the install guide and the HA docker install seems straightforward, but it wasn't clear (to me) about the Supervisor. Is it just another container?

#

I'm aware the Supervisor is optional but it seems like a Good Idea.

red kelp
#

I have HA running in a Proxmox VM. I'm trying to find out how much RAM to give it. From Proxmox point of view, it's using just over 4 of 5GB. But according to HA, Settings, System, Hardware ("Virtual Machine" "ova"), it's only using 1.5GB. Is the extra memory reported by Proxmox hypervisor just the "dirty" memory? Is there any way to make Proxmox think it's using less? I have memory ballooning enabled and as far as I know the QEMU agent is also running. Please 'reply' so I get notified. Thanks for any info!

rough yacht
#

Hi team, I have an networking related question. Background: I have a custom component (mbapi2020) that is in the top100. Since a few days a simple post request from my component to the inofficial Mercedes API fails for a set of users but not all. In 99%-100% of the cases the users are using HAOS. I have rebuild the request for curl and its failing when requested from the homeassistant docker instance. But is successful outside from this environment. Is their anything special like proxies, request-rewriting or header manipulation in the overall HAOS structure for outgoing requests?

slim atlas
#

Afaik HAOS doesn't do anything like that at all

#

We're talking about an integration right?

rough yacht
#

yes right. its an integration

slim atlas
#

And you tried it from the core docker container itself?

#

If you do the same requests in the SSH addon, do you get the same result?

rough yacht
#

yes, i asked the user to connect from the ssh addon with "do exec -it homeassistant bash" to the ha-core and then to execute the command. It failed (answer from external API http400). It was successful on a maschine outside HA, in the same local network. Let me ask one user to run it in the SSH addon again.

slim atlas
#

How many issues did you get?

rough yacht
#

30-40

#

but is a function that is used on the initial onboarding only

#

so it is a lot for 4-6 days

#

however, I assume it is an problem on mb-side... but want to make sure that I do not forget something. I'm a core user and have not that deep knowledge of haos

slim atlas
#

First off, this issue sounds strange. Like if you came in saying, people updated to OS 11.4 and people got problems, i would say, that sounds like a lead

#

And afaik we don't do anything to limit those requests

#

Or modify them

#

Whatever

rough yacht
#

ok, thanks for your support and a great platform. I'll search further

slim atlas
#

So, I'm also leaning towards the mb side

#

Could it be that it's blocking something based on user agent?

rough yacht
#

The user agent is set by the integration

slim atlas
#

Not sure if you sent a user agent with the cURLs, and if different user agents are used by different cURL clients

rough yacht
slim atlas
#

When did you last update the version?

#

Of the user agent?

#

/what's the current version of the app

rough yacht
#

the it would break all their apps. I'm using the same like their official apps - and in generell its working. I have requested 200 pins in the last hours

rough yacht
#

Thanks for your support!

slim atlas
#

Let me know if you find anything, I'm interested

pastel umbra
#

Comparing working / non-working requests could give a hint

rough yacht
#

yes, I even tried cascading proxies to check if someone is manipulating requests. I have an "code"-option in the component to set a "developer" proxy - just for me the see my mistakes early. I tried this with one of the affected users and it was successful. --> So direct traffic was not working, traffic with my proxy worked - this was the reason why I asked this question here...

pastel umbra
pastel umbra
#

one potential difference could be in the ssl communication? tls version used, available crypto ciphers/algorithms/...? curl -vhas details about this:

[...]
* SSL connection using TLSv1.3 / AEAD-CHACHA20-POLY1305-SHA256
[...]
slim atlas
pastel umbra
#

Ah indeed, missed the error code

worthy pelican
#

@dusk phoenix I DMed you for a hopefully reasonable reason (FYI; since I forget whether Discord does something to bury/gate-as-request DMs from non-friends)

Also I am working under the impression you are https://github.com/agners (based on profile pic and similar description on your twitch channel to your Nabu Casa bio)

raven cloud
# red kelp I have HA running in a Proxmox VM. I'm trying to find out how much RAM to give i...

I’m not sure 100% if this is right.. but.. Home Assistant dashboard have under settings->system->hardware a CPU and Memory monitor.. I think you can confirm there if 4Gb are available when 1.5 (37% )are the one in use (according to the virtual machine monitoring) as it is a virtual environment and according to your settings you gave 4/5 available memory (and don’t know how proximos do but VMWare uses sometimes swap memory when the host were it runs do not have space for letting it run) hope this makes sense but again.. it’s just an assumption

crimson dune
#

Hello, I just updated my optiplex to run proxmox. Should I install HAOS or should I install it as a container? is there any benefit of installing as a container?

red kelp
#

Quick tip... some add-ons require (or can greatly benefit from) access to specific CPU instructions which Proxmox might limit the HA VM to. For me, I changed my HA VM processor type from its default of 'kvm64' to 'host' so it would have access to the AVX CPU instruction set which improved the response time (and reduced regular CPU load) of my Whisper speech-to-text add-on.

#

I'm sure I could have found a better way to enable the specifically helpful CPU instruction set(s) sticking with the 'kvm64' processor type, but switching to 'host' was just easier. It does come with some downsides but am ok with it.

#

BTW, I've also used TTEK's Proxmox helper scripts too... including my HA instance.

red kelp
#

If not, is there a single place to list all add-ons' settings including all their CPU and RAM usage?

raven cloud
# red kelp Wait... does the memory reported by HA OS (Settings, System, Hardware) account f...

I think the system monitor is the total system wide CPU and memory of the hardware.. python main process psutil. The Memory and CPU measure of the add on starting from core and supervisor I don’t know how it is calculated.. but is available as supervisor sensor (default are disabled) if the addon reports back to supervisor the memory and cpu used (I guess there is Docker CPU and Memory). There’s a system monitor addon as well that monitors also the CPU temperature on a real.. CPU. Virtual Machines wise depending from the Host the system will be getting a part of the CPU time and Memory but Host OS will also continue to work and need CPU and Memory.

somber lotus
#

Just curious, has anyone heard of home assistant duping itself after a power loss

dusk phoenix
#

What do you mean by "duping"?

somber lotus
# dusk phoenix What do you mean by "duping"?

So I had a power cut for an hour, system came back online and install size was exactly double, checked via putty, nothing unusual, then sftp in to check folder by folder could not find anything. but installed size went from 33 to 66.

#

so I am wondering did something cause the os to double in size (duped itself some how into a separate partitions as I can not find anything to suggest were this double size has come from.

dusk phoenix
#

33 to 66 of what unit? GB?

In general, there is no mechanism built-in which would explain this. So I think this was just a natural grow of the disk image. Are you using virtualization? Also your request is better served in #installation-archived, this channel is about OS development.

somber lotus
raven cloud
#

guys I did just started to work on an issue created by a user that has Proxmox VE with the OS. Checking the logs that he provided me (incredibly grateful for his cooperation) I did notice that somehow the python round didn't work and also it looks somehow there is an issue with AsyncIO.. as per some entries on the logs are a little mixed. I'm writing here because of Proxmox VE uses the OS environment in the VE (the details are in here https://github.com/sca075/valetudo_vacuum_camera/issues/75). Questions is how OS is managing the VE in general and is there any specific OS difference between Proxmox VE and for example PI4 releases of the OS to consider? If there's any docs about this "setup" I missed please let me know.

dusk phoenix
#

The image used on Proxmox VE is typically the ova board (open virtualization appliance). It essentially has some paravirtualization drivers and tools enabled, but that is about it. Compared to a Pi 4 it is mostly the architecture difference (x86-64 vs. aarch64). But otherwise, we try to keep the setups as in sync as possible.

#

I guess you connect to the device via network? I guess I'd check the network setup. Typically Proxmox VE should use bridge networking to make sure HAOS runs directly on the network.

raven cloud
# dusk phoenix I guess you connect to the device via network? I guess I'd check the network set...

I didn’t ask how his device is connected to the network but it is somehow (of course also in my code needs some adjustments) something strange that happens on Proxmox anyway thanks a lot for the clarification, really appreciate it. It’s a kind of challenging period, I hope we will keep up with all the requirements that are daily coming out 😉 hopefully this can be helpful also for you guys, I started to understand how it difficult to keep up with all this platforms..

rough yacht
# slim atlas Let me know if you find anything, I'm interested

Thanks again for your help. I can confirm it is not HAOS related. It looks like that the vendor implemented something like API flow control. My component did not called a unneeded API and as result the vendor API blocked the later calls in the auth flow. I and most of the other users have been not affected as for the client IP correct other traffic was initiated before. As most of the new users are using your great platform, all the error reports stated HAOS... so thanks again for your reaction.

slim atlas
blissful bloom
latent wigeon
blissful bloom
red kelp
#

On HAOS does the Settings->System->Hardware screen RAM usage account for the entire system including all the ADD-ONS too? FYI, I'm getting the same reading from the System Monitor integration's Memory sensors. But this is actually a VM running on Proxmox. Proxmox is reporting most of the allocated memory in-use though. But I thought the QEMU agent was included in HAOS? If it's included, does that mean Proxmox should be more aware of its actual RAM usage? Just trying to understand all the abstraction layers. Thanks for any info!

raven cloud
# red kelp On HAOS does the Settings->System->Hardware screen RAM usage account for the ent...

Apparently in ProxmoxVE HA will react same as it is installed on the VE configured by you.. I think the view of the memory in use (as HA run on a single THREAD) is represented as the total RAM available for the VE and therefore report the value of what is actually in use on that VE. Also if not technically correct, from my experience (developing a custom component) I believe the ADD-ONS memory usage is summed on the total usage. And do not read what is running on Proxmox.. anyhow I think it really depends on how Proxmox configure the VE. (I’m going to test it soon)

red kelp
# raven cloud Apparently in ProxmoxVE HA will react same as it is installed on the VE configur...

I assume when you're at the shell via the console (HA CLI), this is the lowest level you can get to HAOS at the kernel level:

# uname -a
Linux homeassistant 6.1.71-haos SMP PREEMPT_DYNAMIC Tue Jan 9 15:51:13 UTC 2024 x86_64 HAOS
# free --mega
             total    used   free   shared  buff/cache  available
Mem:          4108    1534    489        4        2083       2523```
Settings->System->Hardware is telling me **1.5 GB / 4 GB**, which seems to match above for *used/total*.  So yeah, appears this memory usage value on the HA GUI is the entire system including add-ons.  Now I just wish Proxmox's guest agent (QEMU agent) gave more details... such as a color coded graph of used vs. buff/cache.  Proxmox currently just reports what looks to be used+cache. Maybe one day Proxmox might improve its memory reporting capabilities. 🤞
raven cloud
red kelp
raven cloud
# red kelp Also, doing a ``# ps`` and sorting by memory usage descending shows approximatel...

well I'm not 100% sure but HA use psutil.. so when you see that memory usage.. well, it should be what in reality is used, again for instance reported from the hardware (VE) in this case.. I promise that as soon I have some real hardware (that by the way should be available later today) I can make some test Promox and arrive to some conclusion, for sure the QUEMO agent is calculating it on the real hardware.. and I can just confirm you that psutil is well would do the same.. so.. where is the problem now? VE are a little strange sometimes to understand.. and probably we should move this discussion (I'm not one of the OS developers, but I'm big fun of them 😉 in some other "general" thread.. for instance untill it isn't clear where the problem is. EDIT: and by the way alpine + docker.. probably there is also this to consider..

red kelp
#

@raven cloud I don't think there's any problem/issue going on. First, I was just trying to understand what Settings->System->Hardware screen RAM usage account was accounting for. Now I'm fairly confident it's for HA **and **all the addons too and probably anything else that happens to be running on HAOS.
Second, I was trying to understand Proxmox's VM memory usage view compared to HAOS's memory usage. I'm fairly confident Proxmox's memory view also accounts for Linux OS "cache" that's not necessarily used by any regular processes.
So I think I finally answered my own HAOS-related questions. Have fun on your own Proxmox adventures. 😉 BTW, here's an example pic of Proxmox memory usage that I mentioned earlier:
https://community.home-assistant.io/t/home-assistant-high-memory-usage/180556/24

strange scroll
#

Not sure if this is the right place, but I just wanted to say I updated to HAos 12.0RC1 on my VirtualBox VM under Linux on a Dell laptop, and it just works. Not sure if there is anything I can report except that it starts up normally, memory use and processor seem similar to what I'm used to, very uneventful release for me. Thanks!

reef girder
#

how to enable ssh?

slim atlas
fossil osprey
#

Has anyone been complaining about simple time triggers not firing since 2024.2 release in HAOS installs? Or firing a weird time later? 2hr24m later in my case.

slim atlas
stark cosmos
#

Hey,
few days ago i setup HAOS in Unraid as VM with the 11.5 .qcow2 image:
But it seems that the qemu agent is missing.
Can i install it somehow?

bitter scaffold
knotty cargo
#

Hey, I have a Raspberry Running with HassOS and need to place a additional systemd unit on the root file system. What is the best way to do this?

#

I know this is not intended, but unfortunately there is no alternative. The action must be performed before Homeassistant is started

knotty cargo
#

I found a creative solution. I have created a udev rule that executes a script at startup

dusk phoenix
#

Yeah since the rootfs is largely read only that is probably your only option 😢

Usually the answer is to create an add-on to do what you want it to do.

weak island
#

Was told to give it a try here: If I understood it right, the linux kernel modul for the rtl8xxxu was patched in 6.5 to include a new chipset of that series. Right now it seems that the current kernel is 6.1. Is 6.5 in the pipeline already, or should I try to cherry pick that PR and rebuild the kernel module? I would prefer to just wait a few weeks, before risking to screw up the installation or later on the update path.

warm hawk
#

Hi folks... not sure if this is the right place, but I don't feel like I have enough to log an "Issue".

With HAOS 12.1, my Node-RED instance (self-install on separate machine) started having troubles maintaining its connection -- it ends up in a rapid-fire reconnect cycle. Restarting either Node-RED or HA Core solves it for an hour or so. I've downgraded back to HAOS 12.0 and that resolved the issue.

I don't see anything in the release notes to explain this.

I have HAOS running on ESXi as deployed OVF. Node-RED 3.1.7 is on the same host in a different Debian-based VM.

Any thoughts on how I can help troubleshoot this better or provide a fullsome issue report?

weak island
#

Hi again, after some more research I learned that 12.x already comes with the 6.6 as LTS kernel. However, in the release notes for the Rpi it is mentioned that they still come with 6.1 as the devs follow downstream the buildroot releases and not upstream. Ok, but how can I see if and when 6.6 is in the pipeline?

dusk phoenix
#

(converted to a reply)

dusk phoenix
# weak island Hi again, after some more research I learned that 12.x already comes with the 6....

Yeah at the time we upgraded the other boards, Raspberry Pi did not release a 6.6 officially yet. Just a couple days ago Raspberry Pi OS updated to 6.6, however they seemed to have changed their build system a bit: in the past there was always a tag in their kernel repository for every official release they made. Not this time around.... We'll have to check how they pick releases now, and follow their setup. We first plan 12.2 with a Builldroot update, RPI 6.6 Probably 12.3.

dire ore
#

i built myself a HA OS image locally. how do i OTA it to an existing system?

dire ore
fast arrow
#

Hey! Not sure if this is the right spot but from what I could gather home assistant os support for raspberry pi 5 rests on uboot support. But when I tried looking i to uboot development I was a bit confused by the many forks. So when estimating how far off uboot support for the pi5 is, where so you look?

dusk phoenix
#

HAOS currently supports raspberry pi 5 using the RPi firmware bootflow, without U-Boot. We probably won't switch back to a U-Boot based flow.

There are some patches on the U-Boot mailing list if you're interested in following the progress

gentle frost
#

would it be realistic to build my own operating system for a smart watch where I can fully control my home assistant?

latent wigeon
#

If you can communicate with an HA instance via websockets from your watch, sure

brisk berry
wise bolt
brisk berry
#

I am working on a Mac, I want the image for a raspberry pi 5

dire ore
#

i believe you'll need a linux system. sorry.

wise bolt
# brisk berry I am working on a Mac, I want the image for a raspberry pi 5

Sorry, missed your reply yesterday. Unfortunately mischief is correct, it's not mentioned explicitly in the instructions but the build is expected to be run on a Linux host (or at least a VM). The part creating the data partition (i.e. the hassio package) needs to use loop mounts, that's why the losetup is called in the prepare script.

brisk berry
#

Alright thanks for the help. I’ll try running it on a VM. I do have a pi4, I just thought that it will take way to long to build it on a pi

brisk berry
# wise bolt Sorry, missed your reply yesterday. Unfortunately mischief is correct, it's not ...

Alright so I made a amd64 ubuntu virtual machine on my mac and am able to get through most of the building but I am getting this error:

collect2: error: ld returned 1 exit status
make[4]: *** [../Makerules:533: /build/output/build/glibc-2.38-44-gd37c2b20a4787463d192b32041c3406c2bd91de0/build/math/libm.so] Error 1
make[4]: Leaving directory '/build/output/build/glibc-2.38-44-gd37c2b20a4787463d192b32041c3406c2bd91de0/math'
make[3]: *** [Makefile:484: math/others] Error 2
make[3]: Leaving directory '/build/output/build/glibc-2.38-44-gd37c2b20a4787463d192b32041c3406c2bd91de0'
make[2]: *** [Makefile:9: all] Error 2
make[2]: Leaving directory '/build/output/build/glibc-2.38-44-gd37c2b20a4787463d192b32041c3406c2bd91de0/build'
make[1]: *** [package/pkg-generic.mk:283: /build/output/build/glibc-2.38-44-gd37c2b20a4787463d192b32041c3406c2bd91de0/.stamp_built] Error 2
make[1]: Leaving directory '/build/buildroot'
make: *** [Makefile:33: rpi4_64] Error 2

rapid compass
#

What is the "in view" option in the new update for the WearOS system? It does not refresh the camera live

livid bane
#

Hi Guys, beeing in the HA World quite a while now. I am struggling with a problem:

homeassistant kernel: block nvme0n1: No UUID available providing old NGUID

I installed the HAOS directly on an nvme and attached it to an cm4 on a custom carrier board.

Any thoughts what could cause this problem.
THX and all the best

dusk phoenix
#

Haven't seen this tbh

stable trench
#

Hi All, where is the pip package folder in HAOS ?

dusk phoenix
#

HAOS does not have a Python installation on it's own. Everything is running in containers

warm remnant
#

Hi. I am working on setting up multiple (e.g., 20+) home assistant device hubs based on the Raspberry Pi 5 board, each with a predefined custom configuration. One approach I've considered is setting up a device hub, creating a backup, and then restoring this backup during the initial device hub startup. However, given the large number of device hubs I need to prepare, I'm exploring programmatic solutions. In my first attempt, I built a custom OS image and, during the build process, inserted my base configuration folder onto the hassos-data partitions at /hassos-data/supervisor/homeassistant. I then wrote the image to an NVMe SSD stick. This method seems to work, but there is a lot to Home Assistant (in a good way), and I'm uncertain if this is the best approach—there might be unforeseen side effects. Another method I'm considering involves using a custom supervisor, post-landing page, and employing the restore method from the backups manager. Would anyone have recommendations on how to accomplish this, or any other feedback?

haughty path
#

that sounds interesting, what's the custom configuration that you are attempting to add?

dusk phoenix
# warm remnant Hi. I am working on setting up multiple (e.g., 20+) home assistant device hubs b...

By default we also deploy the landing page. But you can disable that behavior and deploy the latest stable of Core by modifying buildroot-external/package/hassio/hassio.mk. If you want a more controlled environment (version wise), probably using a static/self-maintained stable.json there make sense. But if you want to have full control over the update system etc. a forked version of Supervisor is the way to go yes.

warm remnant
# haughty path that sounds interesting, what's the custom configuration that you are attempting...

Nothing that significant in the sense of what you see others doing. Probably the most significant thing is that all data is being sent to an AWS Postgres server, each device hub has a matching database to write to, and there is a per database user and password that has limited access to only that database. I use the NVMe serial number (which is on the stick label) as the device hub's identifier (using the nvme cli to probe this during startup), so that I can have specific settings for each device hub. Currently I have some custom code in the hassio-supervisor script (/buildroot-external/rootfs-overlay/usr/sbin/hassos-supervisor) that updates the homeassistant configuration (e.g., SUPERVISOR_DATA/homeassistant) based on the NVMe serial number. (prior to starting the supervisor). I'll get this on GitHub shortly.

At the end of the day, this is all about removing as many manual touch points as possible to minimize potential human error, and being as standard as possible for data analysis purposes (as we look at the data across all of the device hubs).

warm remnant
lone sierra
#

Does anyone know if there is a better method that what is described in this post to get access to a python library in Home Assistant? Would it need to be an Add-On, or would an integration be able to do the trick? I'm wanting something a little "cleaner"
https://community.home-assistant.io/t/onkyo-zone-2/14040/15

#

Years ago (before moving and switching from core to HAOS) I was able to just install it on the same UNIX machine and use "simple" command line scripts.

inner widget
#

It seems to be a python library, HA is written in python.
No need to create an add-on, but turn it into an integration instead.

steady bridge
#

@lone sierra There already is an Onkyo integration in Home Assistant. Why not just use it?

lone sierra
slim atlas
#

How good is your eiscp knowledge 😛

steady bridge
#

out of curiosity, what capabilities do you need?

lone sierra
#

Previously, I wrote my own eiscp client using Java, but won’t work on HAOS like it did when I was running HA core.

Looking for capabilities like adjusting inputs, setting FM radio frequency, adjusting volume for center channel for a few.

steady bridge
#

Adjusting inputs should already be possible, is it not working?
You should also be able to change the radio station via presets.
Changing the volume of a particular channel is indeed a very special use case. Probably the best we could do here is provide a special service that allows to make custom calls to the receiver.

lone sierra
velvet sierra
#

Hi, I am trying to add support in hassos for an x86_64 board, that uses coreboot and Seabios. This means I cannot use the (U)EFI approach as used by for the NUC. According to the documentation barebox does support MBR installation. What I am cannot find so far is how I tell barebox in this setting how /where to find the kernel. Does anyone have experience with tihs?

#

It looks like I need to provide the data via the saveenv command. But I am not sure if I can use this during the image build, as it is a barebox buildin command

rare sable
#

Hi jens. We have high restriction to add new boards. However, I'm fine with a generic intel platform for x86_64. But, sorry bios are death since a long time. I know there was some open bios project they invest a half life to such project and don't understand that this kind of technology are gone. A good example is also u-boot they are able to load and execute UEFI applications as opensource. However, there will be no deal to go back to old bios handlings. You can boot an u-boot on that board as first bootloader and boot barebox as second boot loader (like rpi also does). I also hope that I can replace barebox soon with u-boot because of the progress with x86_64 hardware support

#

We use also GPT, and MBR only on boards they have SPL on second sectors what prevent us to use GPT

rare sable
#

@velvet sierra that's not a personal attack or criticism to you

#

A generic system would be amazing but not sure if you will invest time to this if we don't plan to support old bios

velvet sierra
#

@rare sable no worries. no offense taken. 🙂

#

I have been digging a bit more now, and there seems to be a way to use UEFI. It is not enabled by default, by I will try and come back

#

the hardware platform itself I think could be of interest to the HA community. It's the APU2 by pcengines .

rare sable
#

the intel nuc work, so the device would be supported by our kernel, but we don't see the device what I think the redirection need a module?

tropic magnet
#

I'm not sure, but I can definitely test that. Added the issue to my list

rare sable
#

@tropic magnet I created a PR, maybe that fix it. Will create anyway next 2 days v3.8

rare sable
vocal mesa
#

hello/bonjour, any solution to mount a NFS (or any remote FS) on HassOS ? The goal is to use Addons with access to my NAS..... (maybe is not the good channel for this question, sorry)

tropic magnet
#

@vocal mesa That is a support question, please don't use the dev channels for that, try #330990055533576204

rare sable
copper solstice
#

I noticed that by chance yesterday (when checking the changelog for 3.8). Nice!

shy frost
#

Good morning, I have tried to get the RPI-RF-MOD in combination with the HB-RF-USB (converts RPi GPIO to USB) for Homematic to work in a HassOS VM, but I guess it is missing an extra Raw UART kernel module for this. Could you guys point me in the right direction to add this please?

rare sable
#

because HassOS is an embedded system, it need to be exists inside image

shy frost
#

ok, and where would I put the sources I mentioned above?

rare sable
shy frost
copper solstice
#

Following up on my comment in #devs_supervisor-archived might it make sense to allow docker to cache more when SUPERVISOR_DEV is set? Asking here to get more feedback from folks who might have a wider perspective than my own...

rare sable
#

nope, the supervisor is challenge this, not the host

copper solstice
#

Yeah, I don't know why I didn't think of that last night. Must have been too tired 🙂

vocal mesa
#

how to retreive the ingress' url of a addon (something like 192.168.0.31:8123/api/hassio_ingress/SOMERANDOMString76d6s67sd67sd67ss67d67/).... i am able to do it with javascript client side with document.location.... but i need it from a bash file to ajust nginx filter_sub instruction

rare sable
#

@vocal mesa wrong chat

vocal mesa
#

Sorry

rare sable
#

@vocal mesa

vocal mesa
#

Thanks!

shy frost
#

so I got the kernel module for the homematic pcb loaded successfully, but it is not registered as a USB device. It reports to the system as connected via GPIO (just as it would be if it is directly connected to an RPi). I think this would now need an extension of the Homematic CCU addon, because that ones expects a USB device, right?

rare sable
#

Your Buildroot package needs provide udev rules to handling the device correctly and load correct modules. The raspberrymatic use overlays but we work differently (or used at last)

dapper warren
#

Would adding glibc and libstdc++ to the base OS be too much cruft for the sole purpose of enabling VSCode's Remote Development extension?

rare sable
#

there is an add-on for vscode

#

and you can write an docker container (addon) they include what you want for what you want

#

I think you are in the wrong channel @dapper warren because OS is buildroot and did not match with your question

rare sable
#

Release 3.9 is in build pipeline

vocal mesa
#

great, the kernel now include the "snd-usb-audio" module for the nuc version.... do the ova version can have it too ?

rare sable
#

@vocal mesa you can but the CONFIG_SND_USB_AUDIO to device-support.config

rare sable
mild jolt
#

is there some background to why we run on 8123 on hassos? given it's a normal appliance seems it aught to run on standard ports

rare sable
#

Is not part of OS or Supervisor. That is the core default port. You can simple install the nginx addon and run it on 443

copper solstice
#

I kinda agree with you elupus. I personally don't care what port it's on, and have the knowledge and skills to do whatever I want... But what about new less-technically-oriented users? OTOH, I do like the fact that it's non-standard just because it being a PITA to those new users might encourage them to use TLS if they're already switching ports; I have to admit that as user-friendly as running standard on port 80 might be, it wouldn't really be encouraging security on the end-user side.

rare sable
#

please move that discussion to devs_core (why 8123) or devs_supervisor if someone plans to write a certification manager that take care of that for run on 443

#

the only mission of OS is to start a supervisor and provide the env to run it

mild jolt
#

Sure. I just took it up here since for non appliance use it could make sense to be on a nonstandard port. But I just think I'll leave it alone :)

full torrent
rare sable
#

you can make a PR they allow custom firmware on boot partition (/mnt/boot/firmware) and you can copy it there

#

the root file system is RO

full torrent
#

I will see what I can do - but I think the learning curve will be too great!

stone marsh
#

if I add a plugin to docker (on HassOS 3.11), and then a new version of the OS comes out, will that plugin remain, or will I have to make modifications again? (and on that note, how often does the OS get updated?)

rare sable
#

@stone marsh yes it will be removed. A update comes all 2 month and a major release all year, next comes in 1-2 month

#

we do not support any modification that is not documented in the repository. That means, we do not promise that this work over reboots and yes it need at least a rebuild on every update

rare sable
#

maybe it could over life a update, depend on the plugin and what they made on the host system.

stone marsh
#

I'm trying to get the docker-volume-netshare plugin working so I can reference other filesystems on my network. I noticed I am unable to modify the filesystem in the /etc/systemd folder (read-only), so would it be reasonable for me to work on getting this added to future versions?

rare sable
#

No. It's no use case for Home Assistant and we need focus our small team to things they we can long term support and maintain. If there is support for it on supervisor/addon which full use case which working over all our supported platforms, we can look again into this topic

plucky socket
#

@rare sable I get /bin/cli.sh exited 10 when trying to access the host system with login

#

and thrown back to the HA cli login

rare sable
#

is fix with 21

buoyant musk
#

Good morning devs. Is there are document or process for building HassOS? I've figured out why i was having trouble with HassOS, but am having trouble compiling the buildroot barebox stuff to make it permanenent (specifically the script for system0 and system1 needs a minor tweak to mount the volume properly)

rare sable
#

The plan is to migrate it to u-boot. I need retest the current uefi level of u-boot for create an uefi application

#

however, how big is you knowlage about uefi stub?

buoyant musk
#

not very

#

i just know that to get it working for my application (AtomicPi, but would also apply to any machine that treated some storage as "usb"), i need to change the mount command to be (usbdisk|disk)* instead of just disk*

#

it also might fix the nvme issue people are having with NUCs to include the nvme descriptor for the mount point.

#

but if you need a guinea pig for testing uboot for x86_64, i'd be more than happy to help

rare sable
#

there is a folder Documentation which have all kind of docs inside

#

You can also build a u-boot as uefi app/payload and look if you can now boot this on a uefi computer, you need just replace the barebox uefi app with the u-boot one

buoyant musk
#

i'll check it out

rare sable
rare sable
#

Next release will be with buildroot 2020.02.1

#

I update also the gcc version + optimize parameters

rare sable
#

speedup binary

#

we make positive benchmark with our things if we use O3, so we can use it also on the OS now

drowsy shell
#

Can anyone link to current install method for hassos on proxmox VM?

pallid light
drowsy shell
#

Ok I will and thank you. ( I'm asking here because I've heard the devs use proxmox)

lofty harness
#

Please confirm or deny:
an addon can not modify the HassOS kernel.

So a RUN insmod... would be useless in an addon Dockerfile
(maybe I should be posting on #dev_addons ?)
So if I want to use my new ReSpeaker hat, I either change OS to Rasbian or fork and build my own HassOS.

I am running HassOS 3.13 on a Pi 3B.

whole stone
#

it looks like there is a bug in Hass OS 4.5

#

Odroid N2 isn't booting anymore

muted pawn
#

Hi folks, hope everyone is healthy and safe. From my limited understanding of Hassos/Linux, the boot process assumes that the boot device has a particular name, let's say /dev/sda1 or /dev/sda0. I'm trying to install the Hassio NUC image to a NUC-like (small form factor) computer that has eMMC storage. This storage is named /dev/mmcblk1. Having installed many instances of Hassio on small computers, I'm used to transfer the compressed image to an onboard SSD by doing gunzip -c /media/sdc1/hassos_intel-nuc-2.12.img.gz | dd of=/dev/sda bs=4M conv=fsync where the SSD is named /dev/sdaX. This other type of storage has a different name and I guess the reason why the boot process stops in an apparently unrelated step (updating the BT frimware). Is there any way to configure the boot process to allow it to finish with the new storage device name?

rare sable
#

is not needed with UEFI which is used on intel nuc

#

there are a lot (big) documentation about uefi and how block device work inside uefi app

#

the OS itself don't force any device names because we use GPT

muted pawn
#

Thanks @rare sable, the device I'm using is not an actual NUC, it is similar but not the same. Can you point me to a document or website where I can read about the UEFI boot and how to troubleshoot?

rare sable
#

troubleshooting is not needed

#

if that not work, the hardware manifacture did not correct implement the uefi block device handling

#

but you can build a barebox uefi application and load that on you hardware to see what you can read out

muted pawn
#

Thanks Pascal, will look into it right now.

#

By the way, is there anything like a logfile created during the boot process that can help me understand where the actual problem is?

rare sable
#

you can get the log from uefi console

muted pawn
#

Thanks again! will wok on it right now.

rare sable
#

my boards show the uefi console on the first monitor

#

but an uefi board boot also into uefi shell if there is no OS and don't show a output like "no os found"

#

otherwise you board could run a old bios

muted pawn
#

I see. This particular system has only one HDMI port. How do I make the UEFI console to pop up?

rare sable
#

you need ask you hardware manifacture

#

sometimes on the boot option you can jump into uefi shell

#

but if that is an cheap china board, it will not have an uefi and can't boot hassos

muted pawn
#

hmmm... when you say "boot option" you mean the screen that the BIOS displays just before boot allowing you to enter the BIOS, select a boot device?

rare sable
#

bios are from the old days and would replace around 2005 with a new system

#

called uefi

#

but uefi is a lot complexer and expensive which make that manifacture use there old chips and save money

#

so if you have an bios, you don't have an uefi

muted pawn
#

Yes, sorry, did not express clearly. I'll first confirm that this particular model has a UEFI boot. I got two models from the same manufacturer, one based on the celeron N3450, with a 128 GB SSD, which was super easy to setup (I assume it has UEFI) and this board, based on the celeron N3350, that came with eMMC. I'll make sure that the second system has UEFI and will also read the docs you sent me the link for. Is it OK if I bother you if I still have questions?

rare sable
#

sure, feel free to ask on the channel, everybody is welcome

muted pawn
#

Thanks !

rare sable
#

and if the board is UEFI, you need just flash the image to disk

#

if that will not boot, there is an issue on the UEFI implementation which can only fix with a firmware update by manifacture which need include the i.e. eMMC into uefi addressable devices

muted pawn
#

hmmm... not sure if I follow... the eMMC module is soldered to the mainboard, so I can only write an image from a "live" linux distro and them dd the image to the eMMC, no?

rare sable
#

the uefi need read our uefi firmware from the boot partition and we get all devices over the uefi internal block device protocol to find our OS

#

there is no hardcoded devices

#

but if the uefi did not support the disk over his internal block device protocol, we can't find the OS with our bootloader

#

or a bit shorter, if they don't work after you flash our OS to the disk/storage - you can not fix that

#

that means there is an bug on the manifacture uefi implementation

#

however, you can try it. Build you barebox uefi appliation and load it on the board. Now you can play to see what you get - but will be not simple to understand

#

is pretty low level

muted pawn
#

Understood. I just noticed that the computer has two video ports: one HDMI, and one VGA. which port should I expect the uefi console to appear?

#

Just plugged in the computer and went into the Setup Utility. Although it does not explicitly say that it's UEFI, the Boot configuration priority, indicates UEFI: network boot and UEFI: OS, so I guess it should work with Hassio? I also tried booting Hassio entirely from an external USB3.0 SSD flashed with etcher and it works without a hitch

woeful tide
#

Hey 👋 Is there any documentation about modifying/building HassOS with buildroot ?

woeful tide
#

(Probably in the Documentation folder 🙄 nevermind...)

rare sable
#

An important doc change and specification:

rare sable
muted pawn
#

Hi there, @rare sable ! continued debugging the problem I had yesterday with the nuc-like PC. I noticed that the device has UEFI boot properly supported, and the problem is centered on eMMC memory only.
The device boots properly if the Hassio image is written to an actual SSD (the device has eMMC soldered onto the mainboard, but also has a m.2 sata connector). By looking the documentation, just need to confirm with you that the Hassos kernel supports the mmc_block kernel module. Does it?

rare sable
#

I don't know. m2 is supported start with 4.5

#

OS REL-4 RC1 is now in the beta channel

#

happy hacking

muted pawn
#

Thanks @rare sable !

rare sable
muted pawn
#

@rare sable, just wanted to thank you for your work. Been getting immersed on all Hassos stuff, and I have to say that you've put a good amount of love into that!!! thanks for it.

#

... the variety and depth of knowledge that you need to have to put all this together is non-trivial... respect to you, sir!!!

#

Oh, by the way, I read on the NUC.md file on the Hassos git that for installation you recommend: Boot PC to live-environment using PXE or USB
Copy or download the hassos image into your live environment
zcat the image to local harddisk
Reboot

#

the 'zcat' part is what i would like to ask you about: to my knowledge, this command would only show the contents of the image file, but you need to pipe it to something else to actually write it to disk. I normally send the uncompressed data to 'dd'. Do you use something else?

rare sable
#

yes, but I have also an adapter to flash the disk from PC. I would like to have an installer which make that simpler

muted pawn
#

I see... so you zcat the image and then pipe it to dd. I can probably help with the installer. Do you have anything in mind? If you do, I'll give it a quick glance to see if I can build it.

rare sable
#

nothing special. Just somethings that can be booted from a usb stick and allow to flash the image to harddisk

hollow remnant
#

could it not be distributed for the nuc on an install iso?

#

or is that the future plan?

hollow remnant
#

@muted pawn where is the NUC.md file? I am on HAS git but dont see it, which project is that under?

muted pawn
#

Right now we have an image that needs to be moved to the physical device the NUC will use to boot. Some NUCs (and compatibles) have eMMC memory that cannot be taken out of the device (soldered into the board, or maybe the new user doesn't want to open up the device) and for that reason we need to have a program to move the image into the storage device. Right now, you can accomplish this by booting up a Linux distro, and then copy the image into the Storage device... maybe good for you and me, but uber complex for some other users...

#

I was thinking of a bootable image that you can burn with etcher into a USB stick; when it boots up and detects some fixed media (SSD, Rotational drive or eMMC) gives you a little menu to transfer the existing image without you having to open up a shell, looking up devices and issuing the shell commands to uncompress and deploy the image.

#

I can see that Pascal also was in the same line of thinking.

#

Makes sense to you?

hollow remnant
#

thanks altersis, yes i understand now

#

forgive as i havent tried it yet but what file system/structure will you end up with from the image, i ask to make sure it can be expanded to use the entire ssd drive space, or does HA have a utility (since people probably use different size sd cards)

muted pawn
#

The current image has 8 partitions; 7 of them are very small and are used for boot, kernel, and are expected to do minimal writes. The 8th partition is for 'data' (configuration, logs and the home assistant db) so it's expected to be big. As far as I can see, Pascal added a feature on first boot to expand the 'data' partition to occupy the available space on the storage device. If you install the image to a 64 GB SSD, then you'll use around 2 GB for the first seven and then 62 for data. It is also possible to move the 'data' partition to some other device. Don't recall where I saw this explained, but there is documentation in the OS Git about this. If I run into it, I'll paste the link to it in the next couple of hours.

muted pawn
hollow remnant
#

altersis, thanks much!

woeful scroll
#

I have just installed the new qcow2 image to test (kvm/libvirt - no proxmox) - it seems there's no way to log into the machine. No SSH enabled, no console, no serial console. The only way to peak into the hassos vm seems to be via a guest_mount of the image - is that correct or am I missing something? I have two issues where it would be nice to log into the host

pastel umbra
#

I would discuss this in #330944238910963714 maybe, but you can configure spice/vnc output for your VM

#

then you get a prompt where you can use root/login

#

you can configure a serial console from there if that's your preferred way of interacting with the VM

woeful scroll
#

I am using virt-manager for console access and it works fine for other vm's but for hassos I only get output from barebox and the last line is "Loading MS-DOS executable '/mnt/system/bzImage'" - I tried to guestmount the image and as far I can see there's no getty process enabled, hence no console at all. But I most admit that I haven't been a serious Linux user since kernel 2.0, so I might miss something

pastel umbra
#

by 'console', do you mean the 'display/text console/serial console 1' menu entry?

#

or the toplevel 'console' item?

#

in virt-manager

woeful scroll
#

"show graphical console"

pastel umbra
#

and you get barebox output on it?

woeful scroll
pastel umbra
#

I'm surprised you are not getting more

#

in the machine 'details', do you have a display configured?

#

and which image are you using?

woeful scroll
pastel umbra
#

video QXL would be a video card, not a 'display' per se

#

I just downloaded the 4.6 image and booted my preexisting VM using it

woeful scroll
#

I will dig further into this tonight, after work. After I added the display the VM is behaving a bit strange. Need to do some "real" work now - I hope your pull request for serial console gets merged

#

Thank you

terse hazel
#

hope that serial console pull gets merged as well

pastel umbra
#

@terse hazel it was rejected yesterday unless there was more activity today on it

terse hazel
#

Ahhh bummer.

pastel umbra
#

you can try to weight in in that pull request, I did not feel ike fighting for it 😉

terse hazel
#

just took a look. Not going to be figting for it

harsh rain
marble lintel
#

Hi all, i need help... The Home Assistant OS is based on Alpine Linux if i remember correctly... In short if someone want try to help ( party time 😋 ) where to find project spec... Is possible to install the Alpine distro from USB... Why you chosed this distro... ( there are many way to make a personal Linux distro Ubuntu, Open-SUSE ecc )... A simple spec doc ?

#

😎 how deep the white rabbit's den is ?

pastel umbra
#

I guess it fit what the OS devs needed, so they picked that

tropic magnet
#

@fossil merlin This is a development channel, for support please use the other channel that don't have #dev_

#

thanks 👍

fossil merlin
#

sorry I got it

quartz pendant
#

I'm a bit interested in trying to contribute to the underlying os and maybe making it a tad more approachable.
Is there any open RFCs I'm not finding our should I write one before trying to fix and submit patches?

rare sable
#

Bugfix and patches don't need an RFC. The rules for new software and workflow are simple, just need attached to dbus and pickup by supervisor for API/UI and unfortunately if possible also to generic installer

#

And the dev is freezing for new features until we branch REL-4

#

But that should be no blocker for an PR

marble lintel
#

Sry Pascal, the goal is to make a bootable media, that contains "multiple images" and the "possibility of choose" the destination hardware ?

rare sable
#

@quartz pendant architecture repository is the place

#

@marble lintel that's the dev channel not support

marble lintel
#

ok sorry