!QhvgabMQzwEQeWehhZ:lossy.network

NixOS Home Automation

518 Members
Declarative Home Automation and other Sidequests | https://wiki.nixos.org/wiki/Home_Assistant135 Servers

Load older messages


SenderMessageTime
29 Jan 2025
@elonsroadster:matrix.orgcolonelpanicAlso wonder if this warning message in the otbr agent could be relevant: "00:07:33.117 [W] Nat64---------: no mapping found for the IPv4 addres"03:45:55
@robert:funklause.dedotlambdaI've decided against blueprints written in Nix for now: https://github.com/NixOS/nixpkgs/pull/37768804:07:16
@nazarewk:matrix.orgkdnAfter playing with Tuya for a while I want to get started with Home Assistant: got myself HA dongle and a few zigbee-enabled devices. Would you suggest running HA through NixOS or flash RPi4 4GB with their OS?13:11:27
@nazarewk:matrix.orgkdnpersonally I'm more comfortable managing NixOS than whatever distro would be there13:11:56
@laurynasp:matrix.orglaurynaspI guess you are asking for a cons vs pros between the two approaches - I am too new to provide such info 13:57:43
@laurynasp:matrix.orglaurynaspWhat i can share is that my aim is to have as automated setup as i can, hence I am using nixos to run ha in declarative way, and for that reason i picked raspbery pi 4B because it is most compatible. (I am using 8GB rather than 4GB but that is an orthoganal matter). However, for my journal of automated/scripted/decalrative setup of ha, I have been rather disappoitned, because I was not able to find an easy way to automated onboarding, but that is home-assitant's issue, not nixos modules issue. Still, ha module on nixos on pi 4B is the best i was able discover so far. 13:57:58
@laurynasp:matrix.orglaurynasp* What i can share is that my aim is to have as automated setup as i can, hence I am using nixos to run ha in declarative way, and for that reason i picked raspbery pi 4B because it is most compatible with nixos. (I am using 8GB rather than 4GB but that is an orthoganal matter). However, for my journal of automated/scripted/decalrative setup of ha, I have been rather disappoitned, because I was not able to find an easy way to automated onboarding, but that is home-assitant's issue, not nixos modules issue. Still, ha module on nixos on pi 4B is the best i was able discover so far. 13:58:28
@laurynasp:matrix.orglaurynasp* What i can share is that my aim is to have as automated setup as i can, hence I am using nixos to run ha in declarative way, and for that reason i picked raspbery pi 4B because it is most compatible with nixos. (I am using 8GB rather than 4GB but that is an orthoganal matter). However, for my aim of automated/scripted/declarative setup of ha, I have been rather disappoitned, because I was not able to find an easy way to automated onboarding, but that is home-assitant's issue, not nixos modules issue. Still, ha module on nixos on pi 4B is the best i was able discover so far. 13:58:49
@hexa:lossy.network@hexa:lossy.network4 vs 8 GB does matter, if you want the rpi to evaluate and rebuild itself14:00:23
@laurynasp:matrix.orglaurynasp* What i can share is that my aim is to have as automated setup as i can, hence I am using nixos to run ha in declarative way, and for that reason i picked raspbery pi 4B because it is most compatible with nixos. (I am using 8GB rather than 4GB but that is an orthoganal matter). However, for my aim of automated/scripted/declarative setup of ha, I have been rather disappoitned, because I was not able to find an easy way to automated onboarding, but that is home-assitant's issue, not nixos modules issue. Still, ha module on nixos on pi 4B is the best i was able discover so far. My guess is that most ppl workaround to "redeployment" of the configs state is via restoring backups 14:00:25
@hexa:lossy.network@hexa:lossy.network* 4 vs 8 GB does matter, if you want the rpi to evaluate and rebuild itself … aka do automatic upgrades14:00:42
@laurynasp:matrix.orglaurynasp Good point! Hexa, if deploymetns are build remotely, i.e.g by running nixos-rebuild --target-host user@$NIXOS_INSTALL_TARGET_IP switch --flake .#ha-rpi --use-remote-sudo on my main machine, would 4GB still be an issue? 14:03:08
@hexa:lossy.network@hexa:lossy.networkprobably not14:03:26
@hexa:lossy.network@hexa:lossy.networkbut remember that you need an arm64 builder for that, or do cross builds exclusively14:03:46
@laurynasp:matrix.orglaurynaspaha, yet another thing that I forgot to document in my setup - ta14:04:33
@nazarewk:matrix.orgkdnright... I forgot that I wanted to set up mac mini for that...14:15:08
@nazarewk:matrix.orgkdnwould cross-compilation from x86 to rpi4 be significantly longer (I've quite beefy desktop), do you have any materials on how to do that?14:16:07
@hexa:lossy.network@hexa:lossy.networkI haven't14:16:56
@laurynasp:matrix.orglaurynasp

Do you mean "x86_64-linux" or 32-bit system?

My main machine is "x86_64-linux", and on this machine I had add boot.binfmt.emulatedSystems = [ "aarch64-linux" ]; to the config in order to be able to cross-compile for raspberry pi.

In terms of time, first time might take a while (depending on your config could be 10min or whatever), but redeploymetns usually take seconds to a minute, as long as changes don't require the whole univers to be rebuilt.

If it interests you, let me know, and I can share more details

15:15:06
@hexa:lossy.network@hexa:lossy.networkthat is not cross, that is emulation15:15:27
@laurynasp:matrix.orglaurynaspoh... okay15:15:42
@laurynasp:matrix.orglaurynaspwhats the difference?15:15:53
@hexa:lossy.network@hexa:lossy.networkemulation passes native instructions through qemu15:16:14
@hexa:lossy.network@hexa:lossy.networkcross instructs the compiler to compile for another target15:16:25
@hexa:lossy.network@hexa:lossy.networkhttps://wiki.nixos.org/wiki/Cross_Compiling15:16:39
@laurynasp:matrix.orglaurynaspI mean, ultimately that does create binaries locally, and once they are copied to pi, the pi can run15:16:43
@hexa:lossy.network@hexa:lossy.networksetting local and cross systems is the way for that15:16:52
@hexa:lossy.network@hexa:lossy.networkand it is much, much faster than emulation15:17:03
@laurynasp:matrix.orglaurynaspSo it means, that the end results (i.e. binaries) are the same, but the whole process is faster compared to emulation. Or does it mean that the actual binaries produces via emulation would run slower on pi, because those binaries are not as efficient?15:19:24
@mjm:midna.devmjmThe former15:20:01

Show newer messages


Back to Room ListRoom Version: 6