NVMM not working in pair with Tee

[mgerov]

Please provide complete information as applicable to your setup.
• Hardware Platform - Jetson
• DeepStream Version - 6.0
• JetPack Version - 32.4
• Issue Type - bug

I’ve encountered a bug in the usage of tee element with NVMM memory.

The issue arises if I drop frames from one branch of the tee, while processing buffers from the other branch. Subsequently, the other part of the queue sends a timeout signal to the GStreamer bus.

It’s worth noting that when only one of the pipelines (either streaming or DeepStream) is active, everything functions correctly.

Steps to Reproduce:

1. Not working pipeline:
   neural-network-test_not_working.py (2.8 KB)

2. Working pipeline:
   neural-network-test_working.py (3.1 KB)

3. Minimum requirement for replicating the behaviour:
   minimumn_example_not_working.py (1.8 KB)

After dropping a specific number of frames, the described timeout behaviour occurs, leading to the bug I have encountered.

[mgerov]

After further investigation after dropping continuously buffers, the pipeline goes to VOID_PENDING state.
Looks like a deadlock in my opinion.

[Fiona.Chen]

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

@mgerov

The issue arises if I drop frames from one branch of the tee, while processing buffers from the other branch. Subsequently, the other part of the queue sends a timeout signal to the GStreamer bus.

The bufferpool size in nvvideoconvert is only 4 by default. You drop 15 frames per 16 frames, there is no buffer available in the bufferpool. Please enlarge the bufferpool size according to your scenario.
minimumn_example_not_working.py (1.8 KB)