Jetson Nano PR #2548 + RTMP ReoLink GST Pipelines

[cap9qd]

I felt like I was doing wrong cluttering up the PR comments my stuff so I moved the info here. This only works with the PR request from @yury-sannikov here: #2548

ReoLink RTMP pipelines:

    gstreamer:
      inputs:
        - path: rtmp://192.168.1.1/bcs/channel0_main.bcs?channel=0&stream=1&user=frigate&password=TheFrigatePassword123
          raw_pipeline:
            - rtmpsrc location="rtmp://192.168.1.1/bcs/channel0_main.bcs?channel=0&stream=1&user=frigate&password=TheFrigatePassword123" do-timestamp=true timeout=20
            - flvdemux 
            - h264parse config-interval=1
            - nvv4l2decoder enable-max-performance=true
            - nvvidconv
            - videorate 
            - "video/x-raw(memory:NVMM),framerate=5/1"
            - nvvidconv 
            - "video/x-raw, width=(int)640, height=(int)480, format=(string)I420"
            - fdsink
          roles:
            - detect
        - path: rtmp://192.168.1.1/bcs/channel0_main.bcs?channel=0&stream=0&user=frigate&password=TheFrigatePassword123
          raw_pipeline:
            - rtmpsrc location="rtmp://192.168.1.1/bcs/channel0_main.bcs?channel=0&stream=0&user=frigate&password=TheFrigatePassword123" do-timestamp=true timeout=20
            - flvdemux 
            - h264parse config-interval=1
            - splitmuxsink async-handling=true location=/tmp/cache/back-yard-gstsplitmuxchunk-%05d.mp4 max-size-time=10000000000
          roles:
            - record
    detect:
      width: 640
      height: 480

This has the raw pipelines I am using:
frigate.yml.txt

Docker files:
docker-compose.yml.txt
startFrigate.sh.txt

I mistakenly mounted the localtime file from a different docker compose config; if you do this it will mess up the sync between the events DB and the recordings storeage and the web wont be able to find the recordings for the events.

[cap9qd]

I noticed when using the pipeline splitmuxsink the first part of the file name MUST MATCH case and format. I have "Back-Yard" and the location must be "/tmp/cache/Back-Yard-gstsplitmuxchunk-%05d.mp4"

_Back-Yard:
gstreamer:
      inputs:
        - path: rtmp://192.168.1.1/bcs/channel0_main.bcs?channel=0&stream=1&user=frigate&password=TheFrigatePassword123
          raw_pipeline:
...
            - splitmuxsink async-handling=true location=/tmp/cache/Back-Yard-gstsplitmuxchunk-%05d.mp4 max-size-
..._

Also, the pipeline caps for video/x-raw in the detector input section must match the camera resulution and must match the "detector" section resolution.

Back-Yard
  gstreamer:
      inputs:
        - path: rtmp://192.168.1.1/bcs/channel0_main.bcs?channel=0&stream=1&user=frigate&password=TheFrigatePassword123
          raw_pipeline:
           ...
            - "video/x-raw, width=(int)640, height=(int)480, format=(string)I420"
           ...
    detect:
      width: 640
      height: 480

[cap9qd]

JTop:

[cap9qd]

This 4x HD cameras at 2560x1920 @ 5FPS.

[jmorris644]

So I have 5 HD cameras running successfully. All detecting at full resolution. I am not doing any recording yet at this time. Not much CPU usage either. It is getting dark, so I am going to play more tomorrow.

[jmorris644]

I am going to call it a night but I am starting to see some Python bus errors

Fatal Python error: Bus error

[cap9qd]

I am going to call it a night but I am starting to see some Python bus errors

Fatal Python error: Bus error

Did you increase the shm size to match based on the calculation here?

[jmorris644]

Ah, I restarted this process so many times, I forgot. I am not using docker-compose, where do I put it?

[jmorris644]

I tried putting it in /etc/docker/daemon.json and was still getting the errors until I restarted Docker. I am not getting the bus errors anymore.

[cap9qd]

Just an FYI... The dahua camera IPC-HFW1431S-S4 doesn't have audio. My ReoLink cameras do have audio and just assumed...whoops.

[cap9qd]

I am struggling with audio from an RTMP stream. Anyone have any ideas...? I am pretty clueless about GST and only stumbled around with it doing the Jetson tutorials. Any help would be appreciated! The audio isn't a deal breaker but now that I cant make it work I want to!

This is one of my many attempts that has not worked...If I comment out the last 3 lines of the record pipeline the video works.

  Back-Yard:
    gstreamer:
      inputs:
        - path: rtmp://192.168.x.x/bcs/channel0_main.bcs?channel=0&stream=1&user=user&password=password
          raw_pipeline:
            - rtmpsrc location="rtmp://192.168.x.x/bcs/channel0_main.bcs?channel=0&stream=1&user=user&password=password" do-timestamp=true timeout=20
            - flvdemux 
            - h264parse config-interval=1
            - nvv4l2decoder enable-max-performance=true
            - nvvidconv
            - videorate 
            - "video/x-raw(memory:NVMM),framerate=5/1"
            - nvvidconv 
            - "video/x-raw, width=(int)640, height=(int)480, format=(string)I420"
            - fdsink
          roles:
            - detect
        - path: rtmp://192.168.x.x/bcs/channel0_main.bcs?channel=0&stream=0&user=user&password=password
          raw_pipeline:
            - rtmpsrc location="rtmp://192.168.x.x/bcs/channel0_main.bcs?channel=0&stream=0&user=user&password=password" do-timestamp=true timeout=20
            - flvdemux name=demux demux.
            - queue
            - h264parse config-interval=1
            - splitmuxsink name=mux muxer=mp4mux async-handling=true location=/tmp/cache/Back-Yard-gstsplitmuxchunk-%05d.mp4 max-size-time=10000000000
            - queue
            - aacparse
            - mux.audio_0
          roles:
            - record
    detect:
      width: 640
      height: 480
    objects:
      track:
        - person
        - bicycle
        - car
        - motorcycle
        - dog
        - cat
    snapshots:
      enabled: True
    record:
      enabled: True

If I comment out the lines below the video works:

- queue
- aacparse
- mux.audio_0

Here is the stream info from VLC.

[LordNex]

What model of ReoLink are you running. I have 1 RCL520A up currently with 3 more still in boxes waiting for it to be warm enough. The decoding kind sucks though

[cap9qd]

I have ReoLink RLC-410-5MP. I have been using them outside for over a year. In MotionEyeOs/Motion-Project I have been using the RTMP stream with no issues for most of that time. I had the smearing/broken-stream issues until I updated motion and could use the RTMP stream. I dont recall the stream settings for FFMPEG now but it should be doable easily in Frigate 0.9 and this PR.

[jmorris644]

OT: How do I set up notifications just for this discussion? When I enabled 'watch' I get all posts form the complete repository.

[cap9qd]

I am not sure...I just made it and get just the notifications for this discussion and the PR.

Edit: There is a button at the top right of the page for notifications in this discussion:

[jmorris644]

Ah, github had an information comment over it. So I did not see it. In the meantime I created a macro in Outlook to move all github notifications into their own folder. That actually is working well. :)

[jmorris644]

@cap9qd a couple of questions about your config file.

1. Why are you not using any "video/x-rar..." parameters in your Dahua config?
2. Is there a reason for only using a frame rate of 5 for the REO cams and not the Dahua?

[cap9qd]

I used the stock rtsp pipelines for the Dahua camera and not a raw pipeline. I was trying to get the audio into the record file not realizing it didn't have a microphone like the ReoLink.

I have the substream on both Dahua and ReoLink set to 5fps. I copied the config from the PR without changing them much.

I guess the answer is the config were a bit erroneous from piecing it together. Once I have it more stable I'll post the latest configs.

[jmorris644]

I am still getting this error but everything seems to be working correctly. Is this something that I need to fix?

(gst-plugin-scanner:215): GStreamer-WARNING **: 18:02:41.963: Failed to load plugin '/usr/lib/aarch64-linux-gnu/gstreamer-1.0/libgstnvcompositor.so': libgstbadvideo-1.0.so.0: cannot open shared object file: No such file or directory

[cap9qd]

I haven't had any issues with this warning. I think it shows the color bar pattern if the stream/pipeline is bad or unavailable. I haven't been able to figure out how to fix this error yet. If I do I'll post back.

[LordNex]

Looks like it’s missing one of the files. Have you tried a sudo apt update and sudo apt full-upgrade -y yet? On Feb 6, 2022, at 12:04 PM, Joe Morris ***@***.***> wrote:  I am still getting this error but everything seems to be working correctly. Is this something that I need to fix? (gst-plugin-scanner:215): GStreamer-WARNING **: 18:02:41.963: Failed to load plugin '/usr/lib/aarch64-linux-gnu/gstreamer-1.0/libgstnvcompositor.so': libgstbadvideo-1.0.so.0: cannot open shared object file: No such file or directory — Reply to this email directly, view it on GitHub<#2752 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACGXNJG64BKSCC2XCFREIHDUZ2Z2TANCNFSM5NURF7XQ>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jmorris644]

Everything is fully updated and upgraded.

The file is there in /usr/lib/aarch64-linux-gnu/. I have added that directory to LD_LIBRARY_PATH, did an export, and ran ldconfig. It did not help.

[jmorris644]

I am also getting many of these errors:

(gst-discoverer-1.0:272): GStreamer-CRITICAL **: 21:35:24.428: gst_buffer_resize_range: assertion 'bufmax >= bufoffs + offset + size' failed

Any thoughts?

[cap9qd]

I think the dev in the PR is still working on that one. I have it as well without issue. They talked in the pr comments about these errors being erroneous and that adding the jitter buffer helps. I haven't had success getting rid of them with pipeline changes but will keep you posted if I do.

[cap9qd]

I am going to try building the image from the repo and see if some of the things are fixed. I'll keep yall posted...

[yury-sannikov]

Hi @cap9qd, sorry, just seeing this discussion. If you want to build it manually, use this commit I'm updating gstreamer logic to make audio work and some other improvements. It generally works, but I tested it against h265 streams. I believe h264 will not work, except you're using raw_pipeline.

To make audio works, you can try this pipeline (from unit tests)

[cap9qd]

No worries at all. I wanted to move the discussion out of the PR about getting it running and setup so it didn't get too cluttered. I guess I should have asked first... I hope this is okay?

If you want me to test anything against my cameras let me know. They are difficult ReoLink cameras with audio and require the rtmp streams to work reliably I have found. I have two different model Dahua cameras and the other might have audio...

Once I get the build environment up and running I can test whatever commit you'd like.

[yury-sannikov]

(gst-discoverer-1.0:272): GStreamer-CRITICAL

Those errors are coming from gst-discoverer-1.0 a tool to inspect the RTP stream and detect audio and video codecs. The command itself is just as simple as gst-discoverer-1.0 -v $URL. I would ignore all warnings from gst-discoverer although, I'm pretty new to gstreamer and might be wrong.

[cap9qd]

Ah makes sense. Thanks for the heads up.

I am new to gst and was dumping log files and dot files trying to get audio from the ReoLink RTMP stream... Didn't even think to look for an auto detect gst function! Thanks for the tip!

[jmorris644]

FYI: Just in case you run into the error. I had mistyped a username for one of my cameras. What I saw in the Docker log was an ffmpeg error. This is totally understandable, I just wanted to mention it so no one spends too much time looking too long.

[yury-sannikov]

yeah, I need to rename all FFmpeg-related errors to something more neutral or pass the GStreamer instead

[jmorris644]

An interesting tidbit: When I had all of my cameras (5 of them) set at high resolutions the inference speed was between 17 and 25 most of the time. Now that all of the cameras are set at 720 or less the inference speed is higher than 50. I would have thought the opposite.

[jmorris644]

Caution on Docker Upgrade: On the Jetson Nano docker version docker.io_20.10.2-0ubuntu1_18.04.3 will cause issues on the Nano. Errors similar to

docker: Error response from daemon: failed to create shim: OCI runtime create failed: container_linux.go:380: starting container process caused: error adding seccomp filter rule for syscall clone3: permission denied: unknown

If you get these errors you need to downgrade to docker.io_20.10.2-0ubuntu1_18.04.2

It is a known issue with no resolution yet.

[cap9qd]

#2548 (comment)

I need to pin my docker and containerd so I can get the rest of the updates. Or go experimental... But I'll do that on a separate SD card.

[jmorris644]

Here is where I got the info:

https://forums.developer.nvidia.com/t/docker-containers-wont-run-after-recent-apt-get-upgrade/194369

[cap9qd]

Thanks for the info.

[LordNex]

So your running frigate on the Jetson Nano? Do you also have a Jetson Coral? I have DeepStack running on the Nano doing facial recognition. And Frigate in a RPi4 8gig with a Coral TPU. I then use DoubleTake as a add on to Home Assistant to pull it all together with a GUI interface for creating and training new faces and people. Curious how your use case is On Feb 8, 2022, at 10:06 AM, Joe Morris ***@***.***> wrote:  An interesting tidbit: When I had all of my cameras (5 of them) set at high resolutions the inference speed was between 17 and 25 most of the time. Not that all of the cameras are set at 720 or less the inference speed is higher than 50. I would have thought the opposite. — Reply to this email directly, view it on GitHub<#2752 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACGXNJFIQC6ZIRGZBUXR7M3U2E5SBANCNFSM5NURF7XQ>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jmorris644]

I have Frigate running on a jetson nano. I then spun up a 2nd jetson nano and am trying to install deepstate on it. That is when I ran into the docker version issue. I plan to also put Doubletake and Compreface on the 2nd nano. All of this will feed node-red via MQTT. From node-red I can pretty do anything I want. My first attempt will be to have Alexa announce who is at the door.

[LordNex]

I have Frigate running on a jetson nano. I then spun up a 2nd jetson nano and am trying to install deepstate on it. That is when I ran into the docker version issue. I plan to also put Doubletake and Compreface on the 2nd nano. All of this will feed node-red via MQTT. From node-red I can pretty do anything I want. My first attempt will be to have Alexa announce who is at the door.

That the one tool I just haven't played with much, NodeRed. All of my automations I build manually through the Automations page.

I think I'm just going to keep frigate on the Nano with a USB Coral TPU and then use CompreFace as my detector inside HA addon inside my 40 core PowerEdge R620. I'm also running DoubleTake inside HA that pulls all the prices together and gives a very nice interface for training new facial models and keeping them up to date.

At one point I was running 3 detectors. Facebox, DeepStack, and Conpreface and by far ConpreFace was faster, more accurate, and has plugins for age, gender, and if they are wearing a mask.

I'll have to come up with another use for the RPi4B 8gig

[LordNex]

From what I understand and have tried, Frigate can’t utilize the GPU of the Nano. Now that may have changed but the last time I tried I ram into issues as well which is why I just broke down and got the TPU and tied it to a RPi4. Also remember to call the aarch64 version of the docker container in your docker compose file. That and statically setting the OS to boot using the 64 bit version of the OS. I just bought a 40 ***@***.*** 256gig PowerEdge R620 off eBay that eventually have a VM with the TPU passed through. Well technically is a dual processor config with 10 cores each and hyper-threading so a total of 40 manageable CPU cores. I though about looking for a good PCIe TPU/GPU card along with a Haupeg encoder/decoder. But the Haupeg I currently have I don’t think will handle certain digital things. What we need is an all in one hardware solution that does NVR with detection, recognition, and verification, that can handle let’s say 20 streams PoE. It wouldn’t be cheap cause you would have to virtualization inside and would need enough power to handle that and also have 2.5/5GBE interfaces. I wonder if you could use a kubernet cluster of something similar to the Nano. Since the Nano has been declared EOL, the developer of the DeepStack interface has decided to no longer maintain support for it on the Nano so we’re going to have to find another choice or run at a lower software version with possible security issues. On Feb 8, 2022, at 2:09 PM, Joe Morris ***@***.***> wrote:  I have Frigate running on a jetson nano. I then spun up a 2nd jetson nano and am trying to install deepstate on it. That is when I ran into the docker version issue. I plan to also put Doubletake and Compreface on the 2nd nano. All of this will feed node-red via MQTT. From node-red I can pretty do anything I want. My first attempt will be to have Alexa announce who is at the door. — Reply to this email directly, view it on GitHub<#2752 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACGXNJH3JD7X6QKCDZPLPU3U2FZ7HANCNFSM5NURF7XQ>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[cap9qd]

I have it running very well with 5 cameras. One cam is 2MP and the other 4 are 5MP cameras. The detect is working great and the MQTT is also working great. I am not doing any recording yet and I am not using any zones yet. So I cannot attest to those.

I am doing 3x 5mp cameras (ReoLink RTMP stream) and 1x 4mp camera (Dahua RTSP stream). I am using the low res substream for detection (480p) and recording/dumping the high res to disk for recording. I have missed a recording here and there but that's been my only issue. Like Joe I have not used zone yet but do have HA configured and working.

[Elkropac]

Ok looks good. I have 4 cameras and plan to add 2 or 3 more.
But i have just found out, that jetson nano is not available for reasonable price anywhere. I have had it sitting in my shopping cart for about 70$ , i was waiting for frigate to support it ;)
I think, i will probably go with some industrial pc with dual LAN and m.2 connector for coral TPU - but it's not available either

Crazy ...

[jmorris644]

Yeah, Even PIs are going for over $100

[cap9qd]

I am in the same boat. I had the 2GB version of the Jetson "laying around" when MotionEye seems to have died off and wanted a better alternative. So I'll deal with this until I can get another 4gb Jetson or NX and use the 2gb version for deepstack or something like that to play around with. I have a TPU M.2 but it's the wrong key for everything I own now... :( I also have a Coral Dev but I guess frigate support is broken by some issue... Didn't dig too far before finding this and jumping at it.

[LordNex]

Well today is the day. My RPi4 8gig with a tiny 32 gig SD said it was full trying to upgrade from 0.11.0RC2 to stable. So no I'm going to reimage the 4gig Nano with a custom image built for DeepStack, but has everything already enable and built in. So I'll just disable DeepStack. Then use the USB TPU on it for detection. And the GPU for the encoding and decoding. Then have that tied into Home Assistant with DoubleTake and CompreFace as the detector running through MQTT. Im going to use a Class 10 128GB SD Card this time. All my storage goes to my OpenMediaVault NAS running on the PowerEdge. So hopefully when I'm done, it'll run faster and have better and faster encoding and decoding (before there was almost a 1 min lag between what happen on camera and when it was seen on screen, not great for a doorbell cam) and free up a very in demand Raspberry Pi 4B 8gig for something else somewhere down the road.

What Pi based media servers does everyone like. Plex sucked I though and I haven't really found anything else. Music Assistant inside Home Assistant rocks but I was curious

[jmorris644]

@cap9qd Would yo u mind reposting your docker-compose file? I am trying to move to that instead of the CLI and getting errors. I am guessing there are some paste errors in the original file you uploaded? Granted, I am just learning about Docker-Compose.

[cap9qd]

Sure! I have changed a few things. I have moved my config and media to a USB3 case with an SSD and linked it to /opt/frigate for the service.

docker-compose.yml

version: "3.9"
services:
  frigate:
    container_name: frigate
    privileged: true # this may not be necessary for all setups
    restart: unless-stopped
    image: yury1sannikov/frigate.l4t:latest
    shm_size: "256mb" # update for your cameras based on calculation above
    runtime: nvidia
    devices:
      - /dev/bus/usb:/dev/bus/usb # passes the USB Coral, needs to be modified for other versions
    volumes:
      - /opt/frigate/frigate.yml:/config/config.yml:ro
      - /mnt/sda1/frigate/media:/media/frigate
      - type: tmpfs # Optional: 0.5GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 500000000
    ports:
      - "5000:5000"
      - "1935:1935" # RTMP feeds
    environment:
      FRIGATE_RTSP_PASSWORD: "password"
      NVIDIA_VISIBLE_DEVICES: all
      NVIDIA_DRIVER_CAPABILITIES: compute,utility,video

docker-frigate.service

# /etc/systemd/system/docker-frigate.service

[Unit]
Description=Docker Compose Frigate Service
Requires=docker.service
After=docker.service

[Service]
User=frigate
Group=frigate
WorkingDirectory=/opt/frigate
ExecStart=/usr/local/bin/docker-compose up
ExecStop=/usr/local/bin/docker-compose down
TimeoutStartSec=0
Restart=on-failure
StartLimitIntervalSec=60
StartLimitBurst=3

[Install]
WantedBy=multi-user.target

EDIT: I made a user frigate and moved the media to an SSD plugged into the USB3.0 port; I still get slow cache flush warnings on startup. I also linked the docker-compose.yml and frigate.yml from the SSD to /opt/frigate/. Why I did that...I dont know I was going to put the stuff in opt but if the sd crashes I like it being on the SSD.

[jmorris644]

I don't get why I am having so much trouble with the docker-compose.yml file. I am getting this error:

ERROR: The Compose file './docker-compose.yml' is invalid because:
Unsupported config option for services: 'frigate'

Here is my docker-compose file:

version: "3.4"
services:

  frigate:
    container_name: frigate
    privileged: true # this may not be necessary for all setups
    restart: unless-stopped
    #image: blakeblackshear/frigate:<specify_version_tag>
    image: yury1sannikov/frigate.l4t:latest
    shm_size: "350mb" # update for your cameras based on calculation above
    runtime: nvidia
    #devices:
      # - /dev/bus/usb:/dev/bus/usb # passes the USB Coral, needs to be modified for other versions \
    volumes:
      - /home/frigate/config.yml:/config/config.yml:ro
      - /home/frigate/media:/media/frigate
      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "5000:5000"
      - "1935:1935" # RTMP feeds

    environment:
      FRIGATE_RTSP_PASSWORD: "password"
      NVIDIA_VISIBLE_DEVICES: all
      NVIDIA_DRIVER_CAPABILITIES: compute,utility,video

[Elkropac]

I don't get why I am having so much trouble with the docker-compose.yml file. I am getting this error:

ERROR: The Compose file './docker-compose.yml' is invalid because:
Unsupported config option for services: 'frigate'

Here is my docker-compose file:

version: "3.4"
services:

  frigate:
    container_name: frigate
    privileged: true # this may not be necessary for all setups
    restart: unless-stopped
    #image: blakeblackshear/frigate:<specify_version_tag>
    image: yury1sannikov/frigate.l4t:latest
    shm_size: "350mb" # update for your cameras based on calculation above
    runtime: nvidia
    #devices:
      # - /dev/bus/usb:/dev/bus/usb # passes the USB Coral, needs to be modified for other versions \
    volumes:
      - /home/frigate/config.yml:/config/config.yml:ro
      - /home/frigate/media:/media/frigate
      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "5000:5000"
      - "1935:1935" # RTMP feeds

    environment:
      FRIGATE_RTSP_PASSWORD: "password"
      NVIDIA_VISIBLE_DEVICES: all
      NVIDIA_DRIVER_CAPABILITIES: compute,utility,video

Try updating to version: "3.6" , tmpfs size was added in that version

[cap9qd]

I don't know docker and don't know what I'm doing here; I'm a total hack in the most unflattering sense. I basically Google'd "start docker-compose at boot" and this is the solution I found and made work. Maybe you can shed some light on this for me? Where do you put the docker start command and arguments? The only docker experience I have is HA supervised which does all the bits for you... So whoops.

…

On Fri, Feb 11, 2022, 08:08 Joe Morris ***@***.***> wrote: Question, why did you create the systemd service? Doesn't Docker automatically restart the container for you? — Reply to this email directly, view it on GitHub <#2752 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AIBKIUFUH5U4WH7BOZOEDLDU2UJ4BANCNFSM5NURF7XQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.*** com>

[jmorris644]

I am a Docker noob too. So far I am not too happy with my results. :)

But, I did get a Docker-compose file working great on a Ubuntu machine running CompreFace, double-take, and deepstack. I just don;t know why I am having such trouble with Frigate Docker-compose file. Especially that it works for you and not me.

[jmorris644]

Regarding auto-start, I believe the line "restart: unless-stopped" tells Docker to auto start the container on boot. No need for a systemd service.

[cap9qd]

Ah okay. I guess maybe since I am putting that docker compose in a separate file it has issues? Is there a specific place for the docker-compose? Like a /etc/*d/ directory for the....?

[Elkropac]

You can place your compose yaml file anywhere you want and run docker-compose in this directory. It's just a recipe to setup docker container instance. It's not used to run it afterward (only when you want to setup something in different way)

[jmorris644]

Yeah, I have mine in a sub dir of my home directory.

[cap9qd]

I guess you have the EOL right? I am able to test by doing the ole 'docker-compose up' and it works for me...

…

On Fri, Feb 11, 2022, 08:17 Joe Morris ***@***.***> wrote: I don't get why I am having so much trouble with the docker-compose.yml file. I am getting this error: ERROR: The Compose file './docker-compose.yml' is invalid because: Unsupported config option for services: 'frigate' Here is my docker-compose file: version: "3.4" services: frigate: container_name: frigate privileged: true # this may not be necessary for all setups restart: unless-stopped #image: blakeblackshear/frigate:<specify_version_tag> image: yury1sannikov/frigate.l4t:latest shm_size: "350mb" # update for your cameras based on calculation above runtime: nvidia #devices: # - /dev/bus/usb:/dev/bus/usb # passes the USB Coral, needs to be modified for other versions \ volumes: - /home/frigate/config.yml:/config/config.yml:ro - /home/frigate/media:/media/frigate - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear target: /tmp/cache tmpfs: size: 1000000000 ports: - "5000:5000" - "1935:1935" # RTMP feeds environment: FRIGATE_RTSP_PASSWORD: "password" NVIDIA_VISIBLE_DEVICES: all NVIDIA_DRIVER_CAPABILITIES: compute,utility,video — Reply to this email directly, view it on GitHub <#2752 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AIBKIUHYLY5QPR64HXPQYUDU2UK53ANCNFSM5NURF7XQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.*** com>

[jmorris644]

"EOL"? End of line?

[cap9qd]

End of line. Idk I seem to mess that up more than I'd like to admit.

[jmorris644]

What version of docker-compose do you have? I think I have an old one but am struggling how to update it. I have version 1.24

[jmorris644]

I got it working. It was the old version of docker-compose that was causing the issues. I am now at v2.2.3

[cap9qd]

I removed my service and it's all working. Thanks for the help on that!

[jmorris644]

OT: On Sundays in the early AM I have the REO line cameras autoboot. I had created a user just for Frigate with just view access. When I looked at eh REOlink cameras this afternoon I noticed they were all failing. The user that I had set up did not exist after the reboot. Anyone else see this issue?

[cap9qd]

I thought I responded to this...sorry! I have not had any issues with this happening. There is an option in the ReoLink settings for a auto reboot and appears to be configured by default to 2am on Sundays. It might be worth a look to see if it's configured on your cameras. Why it dumped your user since mine apparently have been doing this for years without me knowing (and I have a normal user for video only configured as well that has persisted through reboots).

Edit: Aaaaand I see my response...

[jmorris644]

HAHA, yes, you did respond. Mine reboots every Sunday at 2AM also. But last week I had created the view-only users and Sunday morning the cameras were not working. So in my case, the reboot on Sunday AM eliminated the user somehow.

[yury-sannikov]

Hi folks, I added a GStreamer documentation to my PR. I will be doing a build sometime this week. Please check out the documentation, as your configuration may not be compatible with the new build.

[cap9qd]

Thank you for the update and all your work! My setup has been working great for a few weeks now and the only issues I have had were camera and SSD problems! I really appreciate it!

[cap9qd]

Just had a look at the documentation. Very thorough and will be a lot of help! I stumbled through the gst setup for my cameras and this should help me greatly.

[jmorris644]

My thanks for everything that you are doing. I will take some time today or tomorrow to look at the docs.

[jmorris644]

I see that the main Nano PR got closed. I do not know how this process works. Will a new one get opened for version 11?

[cap9qd]

From the PR comments the PR will be rebased to 0.11 since it seems 0.10 is nearly ready for release. I'd stay tuned...

Edit: sorry I got the two Joe's mixed up... :(

[jmorris644]

I installed a camera directly on the nano yesterday. I cannot figure out the correct gstreamer pipe syntax to make it work. It uses a YUYV format. Help????

ioctl: VIDIOC_ENUM_FMT
        Index       : 0
        Type        : Video Capture
        Pixel Format: 'MJPG' (compressed)
        Name        : Motion-JPEG
                Size: Discrete 1920x1080
                        Interval: Discrete 0.033s (30.000 fps)
                Size: Discrete 1280x720
                        Interval: Discrete 0.017s (60.000 fps)
                Size: Discrete 1024x768
                        Interval: Discrete 0.033s (30.000 fps)
                Size: Discrete 640x480
                        Interval: Discrete 0.008s (120.101 fps)
                Size: Discrete 800x600
                        Interval: Discrete 0.017s (60.000 fps)
                Size: Discrete 1280x1024
                        Interval: Discrete 0.033s (30.000 fps)
                Size: Discrete 320x240
                        Interval: Discrete 0.008s (120.101 fps)
 
        Index       : 1
        Type        : Video Capture
        Pixel Format: 'YUYV'
        Name        : YUYV 4:2:2
                Size: Discrete 1920x1080
                        Interval: Discrete 0.167s (6.000 fps)
                Size: Discrete 1280x720
                        Interval: Discrete 0.111s (9.000 fps)
                Size: Discrete 1024x768
                        Interval: Discrete 0.167s (6.000 fps)
                Size: Discrete 640x480
                        Interval: Discrete 0.033s (30.000 fps)
                Size: Discrete 800x600
                        Interval: Discrete 0.050s (20.000 fps)
                Size: Discrete 1280x1024
                        Interval: Discrete 0.167s (6.000 fps)
                Size: Discrete 320x240
                        Interval: Discrete 0.033s (30.000 fps)
 

[cap9qd]

Managed to the the ReoLink RTMP Stream to mux the aac audio and thought I'd share.

raw_pipeline:
            - rtmpsrc location=rtmp://192.168.x.x/bcs/channel0_main.bcs?channel=0&stream=1&user=user&password=asdf do-timestamp=true timeout=20
            - flvdemux name=demux demux.video
            - queue

            - h264parse config-interval=1
            - splitmuxsink async-finalize=true send-keyframe-requests=true max-size-bytes=0 name=mux muxer=mp4mux location=/tmp/cache/Front-Door-gstsplitmuxchunk-%05d.mp4 max-size-time=10000000000 demux.audio
            - aacparse
            - queue
            - mux.audio_0
          roles:
            - record