[GPUDirect RDMA] Failed to install MLNX-OFED on Cents OS-8.0 with 5.4 Kernel

Infrastructure & Networking Software And Drivers Mellanox OFED
1

I am following below document to install MLNX-OFED package which is required for GPUDirect RDMA

My configuration is -
I have Cent OS-8.0 Install with 5.4.2 Linux Kernel on x86_64 arch.
I have InfiniBand related config flags enable with Kernel. Also install rdma-core separately. Able to validate IB stack & compile ran few CUDA codes.

Able to install CUDA-12.5 toolkit & driver without any issues. I have RTX4000 Ada Gen graphics card install on server for testing.

Getting below error -

image
image1473×742 300 KB

Wondering if we have any way to fix this.

2

Hello,

For CentOS 8.0 I would recommend installing MLNX_OFED’s package via:

Choose the 23.10 LTS version, select the relevant OS, download the .tgz and install using the ./mlnxofedinstall script to install.

If that still does not work, please open a case at enterprisesupport@nvidia.com and it will be handled according to entitlement.

Best Regards,
Jonathan.