Hello, I want to profile vllm in a docker container. I followed the steps in User Guide — nsight-systems 2025.2 documentation and NVIDIA Development Tools Solutions - ERR_NVGPUCTRPERM: Permission issue with Performance Counters | NVIDIA Developer, but still can’t see GPU metrics in Nsight System. I tried some solutions in this forum but didn’t work out. Could anyone help me?
- NVIDIA Nsight Systems version 2024.6.2.225-246235244400v0
- nsys profile --gpu-metrics-devices=help
Possible --gpu-metrics-devices values are:
0: NVIDIA H20 PCI[0000:6b:02.0]
1: NVIDIA H20 PCI[0000:6b:03.0]
2: NVIDIA H20 PCI[0000:69:02.0]
3: NVIDIA H20 PCI[0000:69:03.0]
4: NVIDIA H20 PCI[0000:67:02.0]
5: NVIDIA H20 PCI[0000:67:03.0]
6: NVIDIA H20 PCI[0000:65:02.0]
7: NVIDIA H20 PCI[0000:65:03.0]
all: Select all supported GPUs
cuda-visible: Select GPUs that match CUDA_VISIBLE_DEVICES
none: Disable GPU Metrics [Default]
nsys status -e:
Timestamp counter supported: Yes
CPU Profiling Environment Check
Root privilege: enabled
Linux Kernel Paranoid Level = 4
Linux Distribution = Ubuntu
Linux Kernel Version = 5.15.0-91-generic: OK
Linux perf_event_open syscall available: OK
Sampling trigger event available: OK
Intel(c) Last Branch Record support: Not Available
CPU Profiling Environment (process-tree): OK
CPU Profiling Environment (system-wide): OK
Nsight Systems UI on MacOS: 2025.2.1