Some issues with the recent docker images

[oloukanov]

Hi,

I've tried the last 2-3 Docker images (via Docker Compose), but have some issues/questions.
My camera is PTZ-capable and the firmware is based on thingino-firmware.

• Have problems navigating on the web interface. For example, clicking on Users, then Streams or Recordings, it started spinning, and I had to keep pushing F5 for refresh. Also crashing very often and had to restart the container. Running it on Raspberry Pi 5.
• In Streams, clicking on the pencil to edit the stream does not always work. But the "Delete" button always works. :)
• There are no options to set the ONVIF credentials.
• I don't understand how ONVIF motion events work. In the Recordings, there a several 31-second recordings, although I only selected "motion only".
• Zone editor - "Snapshot unavailable", not sure why.
• PTZ doesn't work. It works in Frigate, for example. Maybe related to ONVIF (see above).
• Live view is not actually "live". I feel like it replays these 31-second segments I mentioned above. And when the container/web interface crashed, the "Live View" was significantly out of time. Refreshing the Live View section is going back in time according to my camera's embedded clock in the stream.
• What is the hamburger menu button in the top-right corner in the Recordings section? Clicking on it has no effect.

Thanks.

[matteius]

This is some great feedback, and aligns with my own experience in some cases, so let me dig in with initial responses:

Have problems navigating on the web interface. For example, clicking on Users, then Streams or Recordings, it started spinning, and I had to keep pushing F5 for refresh. Also crashing very often and had to restart the container. Running it on Raspberry Pi 5.

I run it bare metal on the Pi 5, and on most of my devices, I am not using the docker container much except to debug user issues.

In Streams, clicking on the pencil to edit the stream does not always work. But the "Delete" button always works. :)

I've been adding test CI with playwright that has exposed some of these issues too, Ill cut a new release soon, cannot recall but I think some of those improvements are in main haven't been published. It tends to stem from json response handling/connection close etc from C context, turns out is just harder to get right than in other more web languages. The preferred pattern in the code is to accept the request and offload it to a worker thread for returning the response, to keep the primary web process from getting congested. There have been some past bugs where one or more of the APIs was hogging the whole handler and/or the connection: close was not being well recieved by the browser.

There are no options to set the ONVIF credentials.
I believe there is if you use the Discover via ONVIF flow, but this is an area we can improve for sure -- I guess let me know what you'd like to see out side of the discovery flow.

I don't understand how ONVIF motion events work. In the Recordings, there a several 31-second recordings, although I only selected "motion only".

Motion based recording, and especially buffered events needs a bunch of work still. I finally have some pan tilt cams to play with more, but less time at the moment--would be good to pair with someone on this front. I added initial support but never figured out testing it end to end fully iirc.

Zone editor - "Snapshot unavailable", not sure why.

It queries go2rtc's snapshot end-point, and as of recent changes no longer proxies through lightNVR, it will hit that directly. In the lightnvr fork of go2rtc, I've updated it to not require the ffmpeg binary, and use the lib API calls to get the snapshot, but probably start by inspecting the network tab and seeing where its trying to get the snapshot from and what its returning.

PTZ doesn't work. It works in Frigate, for example. Maybe related to ONVIF (see above).

Similar story to motion events, I am definitely not a 100% onvif expert at this phase.

Live view is not actually "live". I feel like it replays these 31-second segments I mentioned above. And when the container/web interface crashed, the "Live View" was significantly out of time. Refreshing the Live View section is going back in time according to my camera's embedded clock in the stream.

Well there is live view which is webRTC, in my experience it works well or better t han HLS version, which when the stream disconnects from go2rtc, the HLS version isn't recovering well. I recently tried to add some go2rtc health checking to ensure it doesn't go offline, but I am sure there are a multitude of improvements to make around this. I've only see the 30-second looping/stale video on the HLS side, and on webRTC some of my wifi cams come and go, as in they don't always want to establish the webRTC connection -- I haven't gotten to the bottom of this. Sometimes retarting go2rtc/lightNVR helps, sometimes not.

What is the hamburger menu button in the top-right corner in the Recordings section? Clicking on it has no effect.

Great question 😅 -- we can probably remove that, if anyone has good designer experience that could be helpful too--think of it this way, this project has grown both organically and rapidly with the help of AI assisted coding prompts, though I am putting more ephasis on testing framework these days to help the project mature in the scope of using AI tooling to rapidly accelerate progress.

Thanks for your feedback and accurate thoughts -- let me know in what capacities you may be able to assist me in closing the gaps.

[oloukanov]

Hey, thank you for your fast response! Really appreciate it.
I recently revived my old Wyze CamPan1 camera, thanks to the Thingino project, where I've read/found about lightNVR.

I'm not a developer, so sorry if my feedback looks a bit like a waste of time.

I also realize some of the issues presented can be due to my specific configuration and setup. See below.
I use already pre-built Docker images published on ghcr.io.
The /var/lib/lightnvr/data storage I mapped to my NAS share. I tested SMB to the same NAS share but for some reason, the container doesn't like it, so I ended up with NFS connection type. Maybe I can manually mount the local folder on my Pi to my NAS, but I don't think this is the root cause of my other issues.

Also, my camera is on a different VLAN, so I can't use the "discover ONVIF cameras", and Thingino suggests onvif://thingino:thingino@x.x.x.x:80/onvif/device_service for ONVIF URL, so I'm not sure if lightNVR is properly setting it based on the RTSP stream, e.g. rtsp://thingino:thingino@x.x.x.x:554/ch0.
For some reason, I can't use WebRTC, and I'm stuck to HLS only. Tried different browsers. This can explain the Live View loop and missing snapshot in the zone editor, I guess.

About the 30-sec. recording chunks - is it how it works? I was under the impression that the lightNVR only writes the detected and the cache (if the chunks can be named this way) is handled differently.

Thanks again.
Happy holidays.

[matteius]

I think I fixed a number of your issues, 0.17.5 is building now -- all playwright tests are passing and the nav doesn't appear to stall out -- I realized I may not have exposed the go2rtc port in the base docker image, but I believe it is now too.

Disclaimer: I am not saying work is done, there are still some things to iron out including the hamburger menu was apparently a hide/use filters thing that still doesn't work 😅