yayitazale

Hi, you can as we all did, learn reading the docs: https://docs.frigate.video/guides/getting_started https://docs.frigate.video/configuration/ https://docs.frigate.video/configuration/reference PD: as first step, you should reinstall the app from scratch from the App store of unraid, the new version requires a new template.

https://docs.frigate.video/configuration/tls/

This is note related with the template so please post your problem in the github https://github.com/blakeblackshear/frigate/issues

Hi: tmpfs that goes to /dev/shm is what is usable RAM in the host, in your case 7.8GB. If you go to the console of the container and run the same command you will see that is mapped to /tmp/cache tmpfs 954M 9.0M 945M 1% /tmp/cache If you run the "free" command in the container console you will se that you have 1GB of ram. You should read this to solve your issue: https://docs.frigate.video/troubleshooting/recordings

Hi, sorry for the delayed anwser, is ban holiday in Spain. First of all, yes you must reinstall the app from scratch because the new v14 version needs a new template with new entries. Device entries weren't never mandatory entries, and aren't in the new template. All the optional entries have the "remove this if you are not using it" so, as it describes, you should remove them. If the container is no longer available in the frontend of unraid, you should be able to install it again.

If the database update has been made no, there is no way to go back without a backup. Auto updates are very risky.

This is a known problem, and the temperature limits are there to avoid fire (yes, the catch fire). In you image your TPU ins in 33ºC so there is no problem at all, but if you reach the limits I suggest you to add a heatsink (and maybe a fan). Search in internet "heatsink coral tpu" and you will find some 3D printable solutions with 40mm fans.

https://docs.frigate.video/frigate/installation/#calculating-required-shm-size You have to change clicking in "advanced view" of the container and going to the "Extra Parameters:". In V14 I had to increase to double the size of the calculation in order to have stability in the system.

Did you change the SHM size?

Can't help you with the error info.

Please post this errors in github https://github.com/blakeblackshear/frigate/issues You add me to the conversation as I'm expecting same problems (@yayitazale)

You should reinstall the app from the APP store.

No, the database will be adapted but you must check the breaking changes in order to change your config file.

Check records entry: https://docs.frigate.video/configuration/record events: pre_capture: 0 post_capture: 0 required_zones: [] And also https://docs.frigate.video/configuration/zones#restricting-snapshots-to-specific-zones

FRIGATE V14 RELEASED Please read the change-log and the new docs carefully as there are some breaking changes that can affect you. It is recommended to delete the container and install it again from scratch as the new v14 requires a new unraid template to meet the requirements of the new functionalities.

It looks OK but if you want to use hardware acceleration to use less CPU of the host you should check: https://docs.frigate.video/configuration/hardware_acceleration Also, if you want to embed the streams in WebRTC for example in a Home Asssitant frontend, it is a good idea to use the Restream capability: https://docs.frigate.video/configuration/restream

Hi: The webui port change to 8971 it is a breaking change for V14 to add TLS and auth capabilities, and it is already covered in the next template version I sent you in the github: <WebUI>http://[IP]:[PORT:8971]</WebUI> <Config Name="Authenticated UI and API access without TLS. Reverse proxies should use this port." Target="8971" Default="" Mode="tcp" Description="" Type="Port" Display="always" Required="true" Mask="false">8971</Config> The CPU usage seems normal, but you can lower the FPS if you want to have a lower usage. Regarding the config file you can delete the input and output args from each camera if they are the same that you have defined in the global ffmpeg entry. In any case the rest of the config looks ok.

There are some changes that you may need to do. You can see the changes of the next template version here so you edit it and add the changes manually: https://gitlab.com/yayitazale/unraid-templates/-/blob/v14/yayitazale/frigate.xml

go2rtc is not easy to set up

If you can access to the web ui then yes everything is up and running. GPU can be used for hardware acceleration and / or object sound detection: https://docs.frigate.video/configuration/hardware_acceleration https://docs.frigate.video/configuration/object_detectors#nvidia-tensorrt-detector If you already own a TPU, I don't recommend you to use a GPU detector since the TPU is much faster and more energy efficient. For hardware acceleration you can use either the integrated GPU (Intel or AMD) or a dedicated Nvidia card and you will get a substantial CPU usage reduction (the greater the number of cameras, the greater the reduction). It is a little bit tricky to set it up so read de docs and the template instructions carefully.

It is pretty good but you are using go2rtc to create a internal steam, using the restream input profile but not using it since you are pointing to the original camera feed so you are using CPU and GPU for nothing. If you will to use the restream feeds, your camera config part should look like: cameras: Turret: ffmpeg: output_args: record: preset-record-generic-audio-aac inputs: - path: rtsp://127.0.0.1:8554/Turret input_args: preset-rtsp-restream And you should delete the "- rtmp" from the camera roles as it has been replaced in fact by the rtsp restream feeds. In any case, you are using the same input_args and output_args in all the cameras, so you can just add it to the global ffmpeg entry (where you have a input arg that you are not using) like so: ffmpeg: global_args: -hide_banner -loglevel warning -threads 2 hwaccel_args: auto input_args: preset-rtsp-restream output_args: record: preset-record-generic-audio-aac retry_interval: 10 For a better performance, is such a config you will have one thread of ffmpeg to restream and another audo recodification threat to the record role. I suggest you to change the go2trc to this: go2rtc: streams: Turret: - rtsp://192.168.XX.XX:554/11 - "ffmpeg:Turret#audio=aac" PTZ: - rtsp://192.168.XX.XX:554/12 - "ffmpeg:PTZ#audio=aac" and then use this general config for ffmpeg: ffmpeg: global_args: -hide_banner -loglevel warning -threads 2 hwaccel_args: auto input_args: preset-rtsp-restream output_args: record: preset-record-generic-audio-copy retry_interval: 10

Just bind the device to the VM

Run the command "ls -l /dev/" in the unraid console and see if you have the apex_0

Can you post the image of the "Coral Accelerator Module Drivers" plugin?

First of all, is the coral driver finding the coral? Then, can you show how are filling out the template?