[QST]: cufile.log Path Change

[FabianHoerst]

What is your question?
Hello, is there a way to change the cufile.log path? It is always created in the entrypoint folder of my python module.

Thanks in advance!

[jakirkham]

Maybe this doc is helpful?

[mingxin-zheng]

I also noticed this happens with the nvidia pytorch 22.11 container and versions after.
To reproduce it

docker run -it --rm --gpus all nvcr.io/nvidia/pytorch:22.09-py3 bash -c "pip install cucim; python -c 'import cucim'; ls 'cufile.log'"

In 22.10 and earlier:

docker run -it --rm --gpus all nvcr.io/nvidia/pytorch:22.10-py3 bash -c "pip install cucim; python -c 'import cucim'; ls 'cufile.log'"

It will throw the error: ls: cannot access 'cufile.log': No such file or directory

In the cufile.log:

 19-02-2023 06:10:38:318 [pid=97 tid=97] NOTICE  cufio-drv:705 running in compatible mode

I am looking for the correct practices. (1) Is this log notice something that needs fix? (2) If I need to move the log file to other places, is there some convention where this file should belong to?

Thank you

[gigony]

In the https://docs.nvidia.com/gpudirect-storage/troubleshooting-guide/index.html#enable-diff-log-file-app

If the logging:dir property in the default /etc/cufile.json file is not set, by default, the cufile.log file is generated in the current working directory of the application.

Is that the case?

19-02-2023 06:10:38:318 [pid=97 tid=97] NOTICE  cufio-drv:705 running in compatible mode

This means that the system doesn't have a GDS-enabled storage device and drivers, so the GDS API call would execute the regular posix API without the benefit of GDS.

[mingxin-zheng]

Thanks @gigony for the reply.