Skip to content

add cvmfs_config reload step to GPU instructions, and add tip about EESSI-specific config files #685

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ocaisa merged 6 commits into from
Feb 26, 2026
Merged

add cvmfs_config reload step to GPU instructions, and add tip about EESSI-specific config files #685

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
badbb81
add cvmfs_config reload step, add tip about EESSI-specific config files
bedroge Feb 20, 2026
6976d30
add link to CVMFS tutorial
bedroge Feb 20, 2026
4554c09
page -> section
bedroge Feb 20, 2026
befb5e1
remove trailing colon
bedroge Feb 20, 2026
d889e54
fix rendering of bullet list
bedroge Feb 20, 2026
e8cd23e
capitalize first letter
bedroge Feb 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 17 additions & 3 deletions docs/site_specific_config/gpu.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,12 @@ sudo bash -c "echo 'EESSI_NVIDIA_OVERRIDE_DEFAULT=/opt/eessi/nvidia' >> /etc/cvm
sudo bash -c "echo 'EESSI_HOST_INJECTIONS=/desired/path/to/host/injections' >> /etc/cvmfs/default.local"
```

Third, you run the helper script
*Step 3:* To actually reconfigure the variant symlinks, reload the updated CernVM-FS configuration using:
```{ .bash copy }
sudo cvmfs_config reload software.eessi.io
```

*Step 4:* Run the helper script:

```{ .bash .copy }
/cvmfs/software.eessi.io/versions/${EESSI_VERSION}/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh
Expand All @@ -67,6 +72,13 @@ Third, you run the helper script
sudo bash -c "echo 'EESSI_202506_NVIDIA_OVERRIDE=/opt/eessi/2025.06/nvidia' >> /etc/cvmfs/default.local"
```

!!! tip "Use a EESSI-specific CernVM-FS configuration file to configure the variant symlinks"
Instead of using `/etc/cvmfs/default.local` you can also put the symlink configuration in a local configuration file that is specific to the `software.eessi.io` repository
or to the `eessi.io` domain.
For the first you can use `/etc/cvmfs/config.d/software.eessi.io.local`, while for the latter you can use `/etc/cvmfs/domain.d/eessi.io.local`.
Comment thread
bedroge marked this conversation as resolved.
For more details about the different configuration files,
see [the configuration hierarchy section of our CernVM-FS tutorial](https://www.eessi.io/docs/training-events/2025/tutorial-best-practices-cvmfs-hpc/access/client/#configuration_hierarchy).

!!! note "How does EESSI find the linked drivers?"
Copy link
Copy Markdown
Contributor

@boegel boegel Feb 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This should perhaps be a subsection that we can link to, instead of a note.

And perhaps this should go above, even above Configuring runtime support?

Copy link
Copy Markdown
Collaborator Author

@bedroge bedroge Feb 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure, it's quite technical and many users may not really care about how this works. Being able to link to it could be useful though.

Copy link
Copy Markdown
Member

@ocaisa ocaisa Feb 26, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This can be done in a follow-up PR (if it is really necessary). We should get this in so that we can start pointing people towards it.


The runtime linker provided by the EESSI [compatibility layer](../compatibility_layer.md) is configured to search an
Expand All @@ -79,11 +91,13 @@ Third, you run the helper script
If, for some reason, the helper script is unable to locate the drivers on your system you _can_ link them manually.
To do so, grab the list of libraries that need to be symlinked from [here](https://raw.githubusercontent.com/apptainer/apptainer/main/etc/nvliblist.conf).
Then, change to the correct directory:

- For EESSI 2025.06 and later: `/cvmfs/software.eessi.io/versions/${EESSI_VERSION}>/compat/${EESSI_OS_TYPE}/${EESSI_CPU_FAMILY}/lib/nvidia`,
- For EESSI 2023.06: `/cvmfs/software.eessi.io/host_injections/${EESSI_VERSION}/compat/${EESSI_OS_TYPE}/${EESSI_CPU_FAMILY}/lib`

Then, manually create the symlinks for each of the files in the aforementioned list (if they exist on your system) to the current directory.

#### Runtime support when using EESSI in a container: {: #nvidia_eessi_container }
#### Runtime support when using EESSI in a container {: #nvidia_eessi_container }

If you are running your own [Apptainer](https://apptainer.org/)/[Singularity](https://sylabs.io/singularity) container,
it is sufficient to use the [`--nv` option](https://apptainer.org/docs/user/latest/gpu.html#nvidia-gpus-cuda-standard)
Expand Down Expand Up @@ -140,7 +154,7 @@ Thus, you may want to periodically run this script to pick up on new CUDA and cu
$ ls -l /cvmfs/software.eessi.io/versions/2023.06/software/linux/x86_64/amd/zen3/software/CUDA/12.1.1/bin/nvcc
lrwxrwxrwx 1 cvmfs cvmfs 109 Dec 21 14:49 /cvmfs/software.eessi.io/versions/2023.06/software/linux/x86_64/amd/zen3/software/CUDA/12.1.1/bin/nvcc -> /cvmfs/software.eessi.io/host_injections/2023.06/software/linux/x86_64/amd/zen3/software/CUDA/12.1.1/bin/nvcc
```
the `/cvmfs/software.eessi.io/host_injections` directory is special, since it is not part of the actual EESSI repository:
The `/cvmfs/software.eessi.io/host_injections` directory is special, since it is not part of the actual EESSI repository:
it is a CernVM-FS [Variant Symlink](https://cvmfs.readthedocs.io/en/stable/cpt-repo.html#variant-symlinks) that points to
a directory on the local system (`/opt/eessi` by default).
The `install_cuda_and_libraries.sh` script installs CUDA and cuDNN in this local directory, thus un-breaking the symlinks.
Expand Down