Testing and Verification

After the implementation is complete, make sure the GPU is allocated to the Pod by running the following command.

kubectl exec -it h100-rdma-test -- /bin/bash

Once you are inside the container, check the available GPU with the following command.

nvidia-smi

The output should look like this.

Mon Dec 19 10:23:45 2024       
+-----------------------------+----------------------+----------------------+
| NVIDIA-SMI 545.23.08        Driver Version: 545.23.08                    |
+==============================================================================+
| GPU  Name                   Persistence-M | Bus-Id        Disp.A | Volatile |
|                            GPU Memory |
|=============================================================================|
|   0  NVIDIA H100 80GB HBM3  Off  | 00:1E.0     Off |              0 |
| 79993MiB / 81920MiB |
+============================================================================+

If the output shows an H100 GPU with 80 GB of memory, the GPU allocation is successful.