Configuring HPE MLDE for Added GPU Nodes

Describes how to configure HPE MLDE for added GPU nodes in a cluster after cluster expansion.

If you add GPU nodes to the cluster after installing HPE MLDE, you must perform the following steps to ensure HPE MLDE works on these nodes.
  1. Sign in to HPE Ezmeral Unified Analytics Software as an Administrator.
  2. To determine the maximum number of GPUs per node in the cluster, run the following command in the shell with a configured kubeconfig.
    kubectl get nodes -l nvidia.com/gpu.count -o json | jq '.items |
            map(select(.status.capacity."nvidia.com/gpu") | .status.capacity."nvidia.com/gpu" | tonumber) | max // 0'
  3. Click the Tools & Frameworks icon on the left navigation bar.
  4. Navigate to the HPE MLDE tile under the Data Science tab.
  5. On the HPE MLDE tile, click the three-dots button.

  6. Select Configure to open the editor.
  7. Set maxSlotsPerPod to the maximum number of GPUs per node in the cluster.