Failed to open more than 3 nvencoder on a multi-gpu machine

[xiasun]

1. Issue or feature description

Failed to open more than 3 nvencoder on a multi-gpu machine

2. Steps to reproduce the issue

Start an Unreal PixelStreaming demo app inside a docker container, for 4 times.

3. Information to attach (optional if deemed irrelevant)

My machine was equipped with 4 GeForce GTX 1080 Ti, which is suppose to support 3 video encoder each.

I started 4 containers with nvidia-docker, and assign GPU 0, 1, 2, 3 to them respectively, with option -gpus '"device=0"', -gpus '"device=1"', -gpus '"device=2"', -gpus '"device=3"'.
Inside each container, with nvidia-smi, only 1 GPU info is shown.

When trying to start my demo app, the first three container works fine, but the 4th container throw error 10 on nvEncOpenEncodeSessionEx(), which is likely to be the restriction of NVENC.

I wonder whether this behavior is right? Since I was running one container on one GPU only, but the max restriction number of NVENC seems to be 3, for all my GPUs together.

• Some nvidia-container information: nvidia-container-cli -k -d /dev/tty info
• Kernel version from uname -a
• Any relevant kernel output lines from dmesg
• Driver information from nvidia-smi -a
• Docker version from docker version
• NVIDIA packages version from dpkg -l '*nvidia*' or rpm -qa '*nvidia*'
• NVIDIA container library version from nvidia-container-cli -V
• NVIDIA container library logs (see troubleshooting)
• Docker command, image and tag used

[github-actions]

This issue is stale because it has been open 90 days with no activity. This issue will be closed in 30 days unless new comments are made or the stale label is removed. To skip these checks, apply the "lifecycle/frozen" label.

[elezar]

I am closing this issue. If any problems persist using the latest version of the NVIDIA Container Toolkit, please open a new issue.