NVIDIA-container-runtime

Insight & Feedback
1

6.8 0.6 3.82G 88.8M 9760 root 02:38:16 73 0 S 0 0 dockerd --host=fd:// --add-runtime=nvidia=/usr/bin/nvidia-container-runtime

Is there a way to stop above process on my ZImaOS server, I don’t have Nvida GPU, but this run in the background using CPU and causing iowait error along with python.

2026-04-15 21:18:41(+0100) (00:01:28) CRITICAL on CPU_IOWAIT(84.6): dockerd, python3.12, containerd
2026-04-15 21:18:29(+0100) (00:00:06) WARNING on CPU_TOTAL(75)
2026-04-15 21:18:11(+0100) (00:00:06) CRITICAL on CPU_IOWAIT(36): python3, dockerd, python3.12

Also python seems to be a bit old? Should this be updated to 3.14 or 3.15?

2

That dockerd --add-runtime=nvidia=... line is just Docker being started with optional NVIDIA support, it’s not actually using GPU or causing load by itself.

The real issue is your iowait , that usually means a container (likely the Python one) is hammering disk I/O, not the NVIDIA runtime.

I’d check what’s running first:

docker ps
docker stats

Also wouldn’t touch system Python, ZimaOS manages that.

If you share those outputs, we can pinpoint exactly what’s causing it