This is exactly what I came here to say too. Thanks for the tutorial Alan, very helpful as always.

I spent more than i hours trying every biso setting, reinstalling proxmox, trying different hard disk, trying different setting in my routor... and this is what went wrong....

It would be great to mention about new port in text version of the tutorial :)

I'd discovered this too by running 'docker ps', but the response I get from 9443 is "Client sent an HTTP request to an HTTPS server." and no portainer. All other ports don't respond on that IP, just 9443

Ahhh right... I just needed to use IPADDRESS:9443.... okay... all good... moving on.

I'm glad you find them useful! ❤️

Thank you!

Great!

Glad I could help!

There's been a few too many things going on in life recently so I haven't had time for Home Automation and Videos! I will return!

Would be super interested in this!

@@Kasparinow Thanks for the interest. I'll create a sanitised version of my instructions (passwords etc redacted), and post back here when done and where to find it

​@tonymatthews1053 tried today again but these tuts mostly work when the used versions of proxmox docker and so on are same as in the video. Taking the actual versions and something goes wrong a not linux used guy is facing problems.... starts with passthrough of the pcie coral and ends with frigate 0.14 rc1😂

I've not delved into go2rtc yet... if it ain't broke, then don't fix it - my RTSP streams work great!

Glad it was helpful!

yes, had to do the same. It seems frigate needs to write a database into that location now

Thanks so much! and thanks for posting your comment here. I have done several reinstalls without getting Frigate to work. When I removed [:ro] everything started working.

actually you can use command: lsusb

Glad it was helpful, and thanks for the RO/RW tip!

Came looking for this. How can I give rights? I'm not so good with linux. @HomeAutomationGuy maybe this step could be added to the blog?

Glad it helped!

Agree, I am all with Reolink locally

Yup, they process everything local. But you still need to block them from internet, because they phone home.

Came here to say just this.. But the AI locally on the cameras themselves can be inaccurate at times no matter the setting. so im looking into frigate myself to supplement. while the cameras do their own cataloging

@@Sch0field8 Only if you have UID enabled to phone home, the whole purpose of that is to tunnel outside of your network to access the cameras remotely..

i'm using 4 reolink e1 pro 8k cams with shinobi right now and wanted to see if frigate is anything better. The tpu ai stuff confuses me a bit, since reolink PTZ cams, do the autotracking and person/animal/object detection out of the box. so i dont need that coral tpu, do i?? i dont think so, because my proxmox runs on dual xeon with 28cores/56 threads/64GB ram, so the cpu power is no problem at all with 4x 8k cams i have no probs running shinobi on a i7-8700k, what is much slower if frigate does not do the job without the tpu, i am certain i will dump it...

Passing USB's through with connection between port changes can be done in the lxc though. It was (for me, not fluent in linux) a hassle that took hours and all the Google fu I could think of. It boils down to another couple of lines in the lxc config file. And probably some more stuff. Sorry, but I can't remember right now exactly how to do it again. And I certainly don´t remember the sites holding the info. It was several forum posts. In different forums...

Love this reply. Could've been my words hahaha. We all know that feeling@@nilssonrichard

@@BrutalCoding Thanks! In hidsight, I should have documented the steps and procedure. Maybe something else you could have written!? ;)

In this specific case of passing through a Google Coral TPU, using a VM would be worse than a LXC due to how Proxmox passes through usb ports as I've sadly recently found out. I was getting frequent USB crashes when passing through either the coral device or the usb port to a VM, turns out proxmox usb emulation can be problematic at times. Using an LXC solves the problem as it has direct access to the usb controller via the host, added benefit being decreased latency of the coral. It immediately solved my crashing device problem and have lower inference speeds too, would recommend.

You are welcome!

Love it MUCH better than Blueiris!

cheers , was absolutely clueless why it wasn't working

good find

Great tip, thanks for sharing!

No problem!

@@Username-qx9gk good question, two of those is managed through tailscale, as it runs on mobile internet (LTE), no other possibility at location, but tailscale is just for accessing proxmox. I have HASSIO in vms there that are exposed by Cloudflare Argo tunnels, where with Frigate proxy addon I have full access to config etc. in convenient way. 3rd instance is at my home, so fiber connection with OPNsense virtualized to handle routing, reverse proxy and local wildcard certificates. PiKVM is in my wishlist (pricey little thing)

how did you get nvidia to pass from proxmoc>lxc> docker> frigate? i fail to get the drivers on debian working. is there some additional code to be added in the /etc/lxc/xxx? - thanks.

@@cobmahavishnu yeah there is: # Allow cgroup access lxc.cgroup2.devices.allow: c 195:* rwm lxc.cgroup2.devices.allow: c 243:* rwm # Pass through device files lxc.mount.entry: /dev/nvidia0 dev/nvidia0 none bind,optional,create=file lxc.mount.entry: /dev/nvidiactl dev/nvidiactl none bind,optional,create=file lxc.mount.entry: /dev/nvidia-uvm dev/nvidia-uvm none bind,optional,create=file lxc.mount.entry: /dev/nvidia-modeset dev/nvidia-modeset none bind,optional,create=file lxc.mount.entry: /dev/nvidia-uvm-tools dev/nvidia-uvm-tools none bind,optional,create=file Put this in /etc/pve/lxc/xxx.conf also very important, install NVDIA driver on proxmox (you have to build driver, don't install the package), and then in LXC machine do the same build from the same file, but use --no-kernel-modules to have nvidia-smi. Hope that helps :)

@@fizyk1769 thanks - will try. Though sorry for my ignorance what do you mean by building driver on pve and lxc? is there any link/ guide for that?

@@cobmahavishnu hey Mate, sorry for late answer, I thought I hit reply couple of days ago. By building your own driver I meant to download the installer directly from NVIDIA site and running it, you will need couple of dependencies installed (driver will say what it needs). If you need some additional guidance, you will find lot of tutorials on YT with step by step guide :)

I am not familiar with any Frigate LXC, so I'm not sure how this works

I had this eact question too. So from your answer it's just that there are no install methods for Frigate that don't use Docker. Which is kind of funny since Frigate itself can't even know it's in a container, that's point of a container :-) Not a bad answer but somewhat odd, in that we're not getting anything out of this extra layer. I will go read more about Frigate and its installation. Thanks for this video, its production level is top-notch.

No, I run Home Assistant OS inside a VM now. Frigate runs in a separate container on Proxmox and talks to HA and MQTT via the network

@@HomeAutomationGuy Thanks for answering! I’ll look at your other video it probably explains more. I also put all of those files in the /opt/ dir cause of you now 😂

Mine has continued to work after reboots, but maybe I'm just getting lucky!

I will once I figure out how to do it! Seems like Deepstack is no more and I'm not sure where to go with it. Any tips or suggestions?

@@HomeAutomationGuy One of the frigate guys forked it and is maintaining it a bit. Also it is actually a quite simple Programm . I customised it a bit for myself to integrate an alpr… What I am still testing is what detector and model to use. Compreface for example is quite sure of it self. A bit too sure. It frequently reports twigs and other random stuff as known person with 99%. While AI Server is more precise, it seems a bit scared and reports a lot of unknown…

Thank *you*! That's so amazing, I really appreciate this 🥰

you have to shutdown the container, before you issue the mount point command :-)

From the Frigate installation section it has it under devices as - /dev/apex_0:/dev/apex_0 Is that correct?

I don't remember doing that 🤔

Thanks for sharing! Glad you got it up and running

I had the same issue, looks like the script now only sets up the portainer container to forward port 9443, so ipOfYourContainer:9443. You'll have to click around the 'this certificate is busted' notice your browser will show.

150W. But that's running the entire server, not just the video component

Sorry. 150W is for my whole rack, including server, switches, access points, a few small devices, and a UPS

@@HomeAutomationGuy That is not bad. It's like 2 old lightbulbs turned on 24/7!

Interesting, I shall look into the overlay options - thank you!

I use firewall rules to limit the exposure on my network. But your way works just as well, thanks for sharing!

I've taken loads of backups of it. Never tried a restore though 🫣

@@HomeAutomationGuy hmm. I need to look into this more. My snapshot fails to be created.

​@@HomeAutomationGuyI am not trying to be flippant, but until you successfully restore you don't really have it

I don't believe I did. I just piped through the port to the Proxmox container which presumably has the right drivers

@@HomeAutomationGuy thanks, I have no idea why this *!#! USB Accelerator does not work ☺

I am having issues with passing the accelerator too. Frigate fails with `frigate.detectors.plugins.edgetpu_tfl INFO : Attempting to load TPU as usb` and 30 seconds later `frigate.detectors.plugins.edgetpu_tfl ERROR : No EdgeTPU was detected. If you do not have a Coral device yet, you must configure CPU detectors.` With a long stack trace with an exception `ValueError: Failed to load delegate from libedgetpu.so.1.0` I've installed the drivers on Proxmox itself and `lsusb` shows `Bus 002 Device 004: ID 18d1:9302 Google Inc.` on the Proxmox and inside the lxc container. When I tried to do the same from inside the frigate container I see blank next `Bus 002 Device 004: ID 18d1:9302` However what is suspicious `Bus 002 Device 001: ID 1d6b:0003 Linux 6.5.11-8-pve xhci-hcd xHCI Host Controller` inside the frigate docker container and `Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub` on the bus. I installed the driver only to be able to run a pycoral test to see if it was connecting to the host alright and it is working with that test no problem. Any clues to what I am missing?

It turned out that I have `unprivileged: 1` inside the lxc file. After remove that line, everything started to work.

I don't know the answer to that I'm afraid. I've only got the USB coral set up on my system.

Not really, they both work as well as each other. Docker worked perfectly fine, and Proxmox is working perfectly fine. I just really like learning new things, which is why I switched it up when I moved house

@@HomeAutomationGuy Fair enough. Thanks!

I've never really touched it whilst it's been operating so I've not noticed. I've been using it for almost two years now and it has been flawless.

I've heard good things about BlueIris, but I wanted to stick with Linux and that's when I came across Frigate.

Make sure you're using a browser that supports WebRTC

Got it eventually sorted by recreating container. Works pretty well.

Fair points. I made an (uneducated) assumption that the VM would add more overhead than an LXC, but I honestly have no idea either. But I certainly think your method is valid.

@@HomeAutomationGuy a VM does add quite a bit of overhead, simply by being up it will consume more memory than an LXC container (VM has its own kernel end such that need to be in mem!) also an LXC's resource use is more efficiently provisioned/allocated by proxmox than with a VM. Simply because what resources you allocate to an LXC is only a cap to how much it is allowed to use and not an allocation of resources to its exclusive use and the container knows its running on say a 12 core system but only is allowed 2 cores for example. Whereas with a VM the system will "think" that whatever you assigned to it is exclusively available for it. you can ofcourse over provision and use ballooning and such techniques to try and recover that allocation from a vm again but that still is more overhead than native in an LXC container. However not everything is suited for containers, especially from a security and separation standpoint running a full OS in a VM is better and more secure. but as said that does mean you need say a GB of memory for a single app in stead of just just a few hundred megs. as with all things everything is a trade off,

I am having the same problem, did you figure it out?

Sadly not. I don't have any spare M.2 slots left. But if you try it please let me know as I'd love to hear how it went!

On the Frigate documentation website

Great automation! Thanks for sharing!

I actually run both. Unifi protect takes care of my 24/7 recording and their app is perfect for looking at my cameras in real time, their doorbell features, and scrubbing through historical footage. Unfortunately their AI detection isn't great, and frigate does a far better job in that space. So I use Frigate to alert my phone with stills and clips for things that are recognised as important

Ahh makes sense, I was thinking of something similar myself 👍 currently got Asus mesh, and eufy cameras and was thinking of going all in on unify...

@@danjones3956 if you're happy with what you have, then I wouldn't change.

The WiFi seems to drop devices, despite being and expensive router and node with custom firmware. Then eufy is pretty unreliable 😔

Turned out I had a DNS issue, I reinstalled proxmox and set my dns. Then I was able to use helper script.

recreate the image and "repull image" in portainer ui

I don't really know all the ins and outs of Portainer, and just use it to view the stack and check the logs. I'm sure it makes a lot more sense to do it all in Portainer, but I started my Docker journey with the config files and just always stayed there.

Nawww, shucks 😊

Make sure you have your docker compose configuration file in the /opt directory and that it's named correctly

"Laziness" is always a reason 😂 But yes, I agree with you!

I agree, it sure is stupid. But it's the only way I could get it to work

@@HomeAutomationGuy does it have any impact on performance?

You can SSH into the docker PVE and run docker compose pull docker compose up --force-recreate --build -d docker image prune -f

@@HomeAutomationGuy I really appreciate you responding. I usually never get responses, especially on 6 month old videos. Thank you so much!

No problem at all! All part of your friendly neighborhood Home Automation Guy service. I hope it helped!

Because way less control on how you do things.

As @blochtheorem mentioned, it helps give me control and separates my services out

Brilliant! Thanks

Try removing the :ro from the docker compose file mappings

@@HomeAutomationGuy Thanks for getting back to me. I found this same answer in the Installation Frigate Proxmox discussion. I also had errors with it not detecting USB coral device, but I got that figured out as well. I had to reboot my whole proxmox. Everything is running great. Something I have wanted to do for a very long time. I spent 3 days working on this as I am a noob with linux. But pushed through and now I got it. Thanks a million.

@@MikeRoss-zv9nm It took me about the same amount of time as I'm a linux noob too!

Something in Proxmox must have changed as I had Frigate running great in a previous install but this time I get the [Errno 30] Read-only file system: '/config/model_cache error. I removed the :ro but nothing changed. There doesn't seem to be much I can do. The Frigate folks really don't want to support the Docker/Proxmox LXC folks. Their suggestion is put it in Docker (no Proxmox)

Do you mean Frigate instead of Proxmox?

@@HomeAutomationGuy I do, sorry about that. I will edit my comment.

This is the recommended way based on the installation documents. Frigate comes as a Docker container, it doesn't look like there's an easy way to install it outside of Docker.

I don't use the migration or High Availability side of Proxmox, so it's not something I've personally needed to deal with.

@@HomeAutomationGuy in that case I suppose is fine.

If you need to restore you have manual effort again. Recommend doing a vm with this as the passthrough will automatically be restored