jear
jear
Sounds like we can legally start with FROM registry.suse.com/suse/sle15:latest
I am testing this fork https://gitlab.com/fcrozat/nvidia-driver/-/tree/master/sle15 Notes are here https://github.com/jear/nvidia-gpu-operator-sles15-sp3-driver
Operator upgrade is OK, driver update OK. Can you please release the SLES driver image ? Notes are here https://github.com/jear/nvidia-gpu-operator-sles15-sp3-driver
Right, 2 are independant : 1 to adress SLES host, and another to adress RKE2.
I recently have installed RKE2 on Ubuntu and I had to patch again https://github.com/jear/nvidia-gpu-operator-sles15-sp3-driver/blob/main/patch/patch-rke2-nvidia-container-toolkit-daemonset.yaml Regarding SLES related patch (nvidia-driver-daemonset), I suspect nothing has been done.
To me the most important is to have SLES as the underlying OS. When operator detects this OS : 1/ the image is available from registry 2/ the nvidia-driver-daemonset has...
By the way, installing on Ubuntu + RKE1 is OK without patching.
> Not sure if this is the right place but here it goes, I got GPU resources working in k3s running on SUSE 15.4. I had to modify the container...
Yes, in the main OpenCost UI. The current API is able to display gpuCount, gpuHours, gpuCost, gpuCostAdjustment. I guess that gpuCost for a pod is gpuCount*gpuHours. gpuCount will depend of...