Ai_nvr's deepstream is not working properly, dropping and adding streams frequently

When I was testing ai-nvr, deepstream wasn’t working properly, and videos were being added and removed frequently. Here are the steps I took to reproduce it:
1:Delete all containers and restart the machine
2:start nvstreamer。(The 16-way video file has been uploaded before)
3:Follow the Quick Start Guide start jetson-redis(sudo systemctl start jetson-redis)、jetson-ingress(sudo systemctl start jetson-ingress)、ai_nvr(sudo docker compose -f compose_agx.yaml up -d --force-recreate),The VSTS are started together in the ai nvr.
4: add 16-way rtsp stream。
In all started containers ,In the list of all started containers, I am missing sdr-monitoring-controller
sdr-deepstream.log (40.0 MB)
vst.log (5.6 MB)
deepstream.log (6.5 MB)

Seems you are using old version MMJ. Please follow here to use latest JPS release for Jetpack 6.0 GA release: AI NVR — Metropolis on Jetson documentation 0.1.0 documentation

You can check the version of the containers as below image for latest JPS release:

I’ve installed the latest version and now the vst is a separate container. How do I configure more than 8 videos

You can modify the VST config file here: /opt/nvidia/jetson-1.1.0/services/vst/config/vst_config.json on Orin device.

I used nvstreamer on my local machine to simulate the stream deepstream running without problems,

I simulated a live customer stream. The stream was erratic and often disconnected, and deepstream stopped analyzing with an active stream of 0,Look at log, the time is 9:28

I will use nvstreamer to simulate video sources on other machines for testing

deepstream.log (19.8 MB)

I simulated a network disconnection. I stopped nvstreamer container and started nvstreamer container again one minute later. deepstream did not recover and I had to restart vst and ai_nvr to recover

In my case with an unstable rtsp stream, deepstream was unable to reason and had to restart if this was triggering a known problem with deepstream (DeepStream FPS degradation). Unstable streams can occur where there are connections but no frames for a few seconds.

Can you recover it by restart DeepStream: $ docker restart deepstream?

I recovered it by restarting vst and ai_nvr. Let me add one more piece of information: deepstream’s log stopped printing when the problem occurred

I summarize the following phenomena:

1: Use nvstreamer as the video source, stop the nvstreamer container directly, in this case, restart vst and deepstream

2: When using unstable stream, deepstream has no active stream, but deepstream log is still printing, in this case, you can directly restart vst (my understanding is that vst has been disconnected from the video source).

3: When using unstable streams, deepstream has no active streams, and all logs are printed. At this time, it is necessary to restart vst and deepstream at the same time to recover

Please refer this topic: Testing ai-nvr adds 16-way video streams and deepstream does not perform inference - #4 by kesong