[Regression] nvmultiurisrcbin does not trigger RTSP reconnect monitoring on DS 9.0/9.1 x86

Hello, sorry to disturb you

Hardware Platform: x86 dGPU
GPU: Tesla T4
NVIDIA Driver Version: 590.48.01
Operating System: Ubuntu 24.04.4 LTS
DeepStream Version: 9.0 and 9.1
Application: Python / GStreamer

I configured nvmultiurisrcbin with:

rtsp-reconnect-interval=10
rtsp-reconnect-attempts=-1

When an RTSP source disconnects, nvmultiurisrcbin does not reconnect.
The application bus also does not receive the expected warning:

No data from source since last ... Trying reconnection

After enabling GStreamer debug logs, only the underlying rtspsrc
disconnection can be seen:

WARNING from src: Could not read from resource.
Debug info: ../gst/rtsp/gstrtspsrc.c(5993):
gst_rtspsrc_loop_interleaved ():
/GstPipeline:multi-rtsp-pipeline/GstDsNvMultiUriBin:src/
GstBin:src_creator/GstDsNvUriSrcBin:dsnvurisrcbin0/
GstRTSPSrc:src:
The server closed the connection.

However, GstDsNvUriSrcBin does not print “No data from source since
last”, “Trying reconnection”, or “Resetting source”.

Test results on the same x86 server:

Environment / component Result
DS 9.0 standalone nvurisrcbin Works
DS 9.0 nvmultiurisrcbin with static uri-list Fails
DS 9.0 nvmultiurisrcbin with REST-added source Fails
DS 9.1 nvmultiurisrcbin Fails
DS 7.1 x86 nvmultiurisrcbin Works

The same RTSP source and the same reconnect properties were used in
these tests.

This does not appear to be specific to REST dynamic source creation,
because static uri-list sources also fail on DS 9.0.

It appears that the internal GstDsNvUriSrcBin created by
nvmultiurisrcbin does not start its reconnect monitoring on
DeepStream 9.0 and 9.1 x86, while standalone nvurisrcbin works.

Is this a known regression in nvmultiurisrcbin?
Is there a patch or additional property required to enable its internal
RTSP reconnect monitoring?

If the current version cannot be resolved, I hope it can be fixed. Thank you!

@Morganh
Hello, could you please take a look? Thank you!

Moving to Deepstream forum since it is related to deepstream.

I am checking. are you testing deepstream-app with virtual camera? if so, how did you simulate “The server closed the connection” of " RTSP source disconnects"?

nvmultiurisrcbin switches the reconnect interval to init-rtsp-reconnect-interval after an RTSP warning/error. Since init-rtsp-reconnect-interval defaults to 0, setting only rtsp-reconnect-interval=10 can reset the active reconnect interval to 0 and disable reconnect monitoring.
you can set “init-rtsp-reconnect-interval=5” . Please try the following workaround.

gst-launch-1.0 nvmultiurisrcbin uri-list=rtsp://127.0.0.1:8554/stream0 width=1920 height=720 rtsp-reconnect-interval=10 init-rtsp-reconnect-interval=5 rtsp-reconnect-attempts=-1 ! fakesink

Set “init-rtsp-reconnect-interval=5”, and it reconnected normally. Thank you very much! Was it originally in rtsp-reconnect-interval, and now it has been split into two parameters? One is the reconnection waiting time, which is the interval between each reconnection, and the other is how long it takes to determine a disconnection if no data is received