Inference stops after some time when video storage is enabled in VST

Accelerated Computing Intelligent Video Analytics Metropolis Microservices for Jetson
, ,
1

Description:
I’m experiencing an issue where, after a period of normal operation with VST and DeepStream running two RTSP video sources, the inference process stops unexpectedly and object detection ceases.

Details:
When I disable jetson-storage and configure VST not to record video streams, the inference and detection flow remain stable over time with no interruptions. This suggests that the issue might be related to how VST manages the GStreamer pipeline between nvinfer and the video recording module.

Question:
Could this be a known issue related to the interaction between inference and video storage in VST?
Is there any recommended configuration or workaround to keep video recording enabled without affecting inference stability?

Environment:

  • Platform: Jetson Orin NX - L4T 36.4.3

  • VST version 1.3.0-24.07.1-aarch64

  • DeepStream version: 7.1-public-v1

  • Number of video sources: 2 RTSP streams

4

Please have a try to set the max video storage size based on here: Storage Config — Jetson Platform Services documentation

5

The current value I have configured in VST for total_video_storage_size_MB is 1000000 MB. Why would setting it to 10000 MB prevent the system from failing?

6

Do you mean the issue has gone after setting total_video_storage_size_MB to10000 MB ?

7

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks.