In ubuntu server, do I have to use nvidia docker for nsight compute?

Developer Tools Nsight Compute
1

Hi! I am using nsight compute on ubuntu server, without docker, it shows that:

a100-01% ncu python test_delete.py
==PROF== Connected to process 908665 (/home/zyhuang/miniconda3/envs/py_hzy_new/bin/python3.11)

==ERROR== Failed to open/create lock file /tmp/nsight-compute-lock. Please check that this process has write permissions on this file. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.

==ERROR== Failed to open/create lock file /tmp/nsight-compute-lock. Please check that this process has write permissions on this file. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
==ERROR== Failed to profile kernel "distribution_elementwise_grid..." in process 908665
==PROF== Trying to shutdown target application
==ERROR== The application returned an error code (9).
==ERROR== An error occurred while trying to profile.
==WARNING== No kernels were profiled.
==WARNING== Profiling kernels launched by child processes requires the --target-processes all option.

Maybe I have to use docker? As root? But you know, in docker, for the purpose to keep status the same to the host machine, we need to add user (which is not root) (so we can change files, and the file can be changed outside). I have tested, if I use root in docker, I can run nsight compute, but not root, no. So it is very hard to change files. I have to use vim, but no vscode, also when I shut down the docker, next time even as root, I can not change the files… Very complicated…

Could someone kindly help me? Thank you!!

2

I see that you also submitted Is it necessary to use docker for nsight compute? It’s likely that the advice there could help here as well. Can we handle the issue on that thread or is there another issue related to this thread?

1 Like