nvh264enc no longer works after upgrading to 22.04

[ghost]

The 22.04 docker image show black screen (no video) in both Chrome or Firefox.
The web GUI works but the HTTP server seems to suck if there are more than one browsers connecting. (the second connection will time out)

I am using the docker image sha256:23bb0ceae21a6cd0e89f1441bb8d12f751276c653db3c67da2d33b8e809ea770 from 6c6dd17

[ehfd]

The web GUI works but the HTTP server seems to suck if there are more than one browsers connecting. (the second connection will time out)

The first one works then?

We don't support more than one browser (although planned), unfortunately. Use noVNC for that, please...

Duplicate of selkies-project/selkies-gstreamer#39

[ghost]

The first one works then?

No. The 22.04 docker image show black screen (Waiting for video stream.) in both Chrome or Firefox.
The logs shows [webrtc] [ERROR] attempt to send data channel message before channel was open.

[ehfd]

Do you have a TURN server?

[ehfd]

If not, try https://www.metered.ca/tools/openrelay/.

[ghost]

No. but 0e61177 was working without a TURN server.
also the docker is in the same local network as my machine.

[ghost]

I have set up the TURN server but the error persists.

I changed the WEBRTC_ENCODER from nvh264enc to vp8enc and it works now, but the encoding is super slow.

I am using a desktop 3090 GPU, and nvh264enc was working before.

[ehfd]

@nick008a When the error happens, please paste the /tmp/selkies-gstreamer-(randomhash).log file

[ehfd]

The other two log files entrypoint-* and pulseaudio-*, please. @nick008a

[ghost]

Failed to init cuda, ret: 0x324 seems to be suspicious. nvidia-smi shows the GPU.

[ehfd]

@nick008a Unfortunately, I think this is something with your GPU. Could you reinstall your host GPU, then reinstall the nvidia-container-toolkit too?
First could you try again with x264enc too?

[ghost]

x264enc works on both Firefox and Chrome.
I have just reinstalled nvidia-container-toolkit=1.10.0-1 and nvidia-dkms-470=470.161.03-0ubuntu0.22.04.1 but the problem persists.

[ehfd]

Maybe you should upgrade the driver version. The internal CUDA version was upgraded to 11.7 because that is the minimum required for 22.04.

[ghost]

solved by upgrading nv driver