RHEL 8 x86_64 repository wrong package file size about cuda-repo-management

[kmittman]

Hi @pjgeorg I'm taking a look. Preliminary check does seem to a metadata and/or CDN issue, so moving from yum-packaging-precompiled-kmod to cuda-repo-management git repository for tracking.

[kmittman]

The root cause was some incorrect metadata introduced last month during the 418.226.00 posting.

rhel8/x86_64/cuda-compat-10-1-418.226.00-1.x86_64.rpm
rhel8/x86_64/cuda-drivers-418.226.00-1.x86_64.rpm
rhel8/x86_64/kmod-nvidia-latest-dkms-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-NVML-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-NvFBCOpenGL-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-cuda-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-cuda-libs-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-devel-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-driver-libs-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-kmod-common-418.226.00-1.el8.noarch.rpm
rhel8/x86_64/nvidia-libXNVCtrl-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-libXNVCtrl-devel-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-modprobe-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-persistenced-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-settings-418.226.00-1.el8.x86_64.rpm
rhel8/x86_64/nvidia-xconfig-418.226.00-1.el8.x86_64.rpm

We use the --update flag with createrepo_c generate the metadata which continued to propagate this issue in subsequent releases.

I have re-generated the RPM metadata from scratch for the rhel8/x86_64 repository and validated the rest of the CUDA repos.

[kmittman]

Thank you for reporting this issue @pjgeorg please let me know if see any similar mismatches in the future, closing as resolved.

• ✔️ $ sudo dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda- rhel8.repo
• ✔️ $ sudo dnf install cuda-compat-10-1
• ✔️ $ sudo dnf module install nvidia-driver:418

This was not the usual issue seen in #1, so I have added a new script repo-validate.sh to check for this going forward.

$ ./repo-validate.sh --mirror=/path/to/snapshot --distro=rhel8 --arch=x86_64 
[...]
cuda-compat-10-1-418.226.00-1.x86_64.rpm [5713804] [b6556b47df]
cuda-drivers-418.226.00-1.x86_64.rpm [7156] [066f1bbb7e]
kmod-nvidia-418.226.00-4.18.0-305.19.1-418.226.00-3.el8_4.x86_64.rpm [12744460] [5578677cfa]
kmod-nvidia-418.226.00-4.18.0-305.25.1-418.226.00-3.el8_4.x86_64.rpm [12744408] [f0e1f9e42d]
kmod-nvidia-418.226.00-4.18.0-348-418.226.00-3.el8.x86_64.rpm [12744232] [94a2644460]
kmod-nvidia-418.226.00-4.18.0-348.2.1-418.226.00-3.el8_5.x86_64.rpm [12744480] [b68c1c6201]
kmod-nvidia-latest-dkms-418.226.00-1.el8.x86_64.rpm [12394276] [ad19aa551b]
nvidia-driver-418.226.00-1.el8.x86_64.rpm [2590320] [803f8b814b]
nvidia-driver-NVML-418.226.00-1.el8.x86_64.rpm [474164] [6a2f622370]
nvidia-driver-NvFBCOpenGL-418.226.00-1.el8.x86_64.rpm [112140] [4d8b997c73]
nvidia-driver-cuda-418.226.00-1.el8.x86_64.rpm [309724] [a165bfb9ec]
nvidia-driver-cuda-libs-418.226.00-1.el8.x86_64.rpm [24639824] [752f329c15]
nvidia-driver-devel-418.226.00-1.el8.x86_64.rpm [12508] [5067b47fdb]
nvidia-driver-libs-418.226.00-1.el8.x86_64.rpm [35801420] [1274ad473b]
nvidia-kmod-common-418.226.00-1.el8.noarch.rpm [10272] [35a7931edd]
nvidia-libXNVCtrl-418.226.00-1.el8.x86_64.rpm [51968] [208d000da8]
nvidia-libXNVCtrl-devel-418.226.00-1.el8.x86_64.rpm [55280] [942d4865b7]
nvidia-modprobe-418.226.00-1.el8.x86_64.rpm [74256] [ba3bc5b729]
nvidia-persistenced-418.226.00-1.el8.x86_64.rpm [106876] [f669939931]
nvidia-settings-418.226.00-1.el8.x86_64.rpm [3645528] [5a9a8f551d]
nvidia-xconfig-418.226.00-1.el8.x86_64.rpm [269540] [577e2cee84]
[...]

Everything matches now.