使用gpu-manager调度的pod起不来，会报Segmentation fault (core dumped) ，cuda版本是11.1

[Justin-ZL]

No description provided.

[DennisYoung96]

so which gpu card u used?

[ls-2018]

I had the same problem
tesla p4

[hnyoumfk]

Maybe your application call nvmlInit() to init CUDA enviroment.

Try add this code at the start of your python entrypoint to find out which function makes seg fault. ( if your application is writen in python)

import faulthandler
faulthandler.enable()

And if nvmlInit() make this happen, these code can make it througth

from ctypes import CDLL, c_int, byref
nvml_h = CDLL("libnvidia-ml.so.1")
nvml_h.nvmlInit_v2()