Setup
Suggest changes
PDFs
This section describes the initial setup tasks that need to be performed in order to utilize NVIDIA NGC enterprise software within an NVIDIA AI Enterprise environment.
Prerequisites
Before you perform the steps that are outlined in this section, we assume that you have already deployed the NVIDIA AI Entrprise host software by following the instructions outlined on the Initial Setup page.
Create an Ubuntu Guest VM with vGPU
First, you must create an Ubuntu 20.04 guest VM with vGPU. To create an Ubuntu 20.04 guest VM with vGPU, follow the instructions outline in the NVIDIA AI Enterprise Deployment Guide.
Download and Install NVIDIA Guest Software
Next, you must install the required NVIDIA guest software within the guest VM that you created in the previous step. To download and install the required NVIDIA guest software within the guest VM, follow the instructions outlined in sections 5.1-5.4 in the NVIDIA AI Enterprise Quick Start Guide.
|
When performing the verification tasks outlined in section 5.4, you may need to use a different CUDA container image version tag as the CUDA container image has been updated since the writing of the guide. In our validation, we used 'nvidia/cuda:11.0.3-base-ubuntu20.04'. |
Download AI/Analytics Framework Container(s)
Next, you must download needed AI or analytics framework container images from NVIDIA NGC so that they will be available within your guest VM. To download framework containers within the guest VM, follow the instructions outlined in the NVIDIA AI Enterprise Deployment Guide.
Install and Configure the NetApp DataOps Toolkit
Next, you must install the NetApp DataOps Toolkit for Traditional Environemnts within the guest VM. The NetApp DataOps Toolkit can be used to manage scale-out data volumes on your ONTAP system directly from the terminal within the guest VM. To install the NetApp DataOps Toolkit within the guest VM, perform the following tasks.
-
Install pip.
$ sudo apt update $ sudo apt install python3-pip $ python3 -m pip install netapp-dataops-traditional
-
Log out of the guest VM terminal and then log back in.
-
Configure the NetApp DataOps Toolkit. In order to complete this step, you will need API access details for your ONTAP system. You may need to obtain these from your storage admin.
$ netapp_dataops_cli.py config Enter ONTAP management LIF hostname or IP address (Recommendation: Use SVM management interface): 172.22.10.10 Enter SVM (Storage VM) name: NVAIE-client Enter SVM NFS data LIF hostname or IP address: 172.22.13.151 Enter default volume type to use when creating new volumes (flexgroup/flexvol) [flexgroup]: Enter export policy to use by default when creating new volumes [default]: Enter snapshot policy to use by default when creating new volumes [none]: Enter unix filesystem user id (uid) to apply by default when creating new volumes (ex. '0' for root user) [0]: Enter unix filesystem group id (gid) to apply by default when creating new volumes (ex. '0' for root group) [0]: Enter unix filesystem permissions to apply by default when creating new volumes (ex. '0777' for full read/write permissions for all users and groups) [0777]: Enter aggregate to use by default when creating new FlexVol volumes: aff_a400_01_NVME_SSD_1 Enter ONTAP API username (Recommendation: Use SVM account): admin Enter ONTAP API password (Recommendation: Use SVM account): Verify SSL certificate when calling ONTAP API (true/false): false Do you intend to use this toolkit to trigger BlueXP Copy and Sync operations? (yes/no): no Do you intend to use this toolkit to push/pull from S3? (yes/no): no Created config file: '/home/user/.netapp_dataops/config.json'.
Create a Guest VM template
Lastly, you must create a VM template based on your guest VM. You will be able to use this template to quickly create guest VMs for utilizing NVIDIA NGC software.
To create a VM template based on your guest VM, log into VMware vSphere, righ-click on the guest VM name, choose 'Clone', choose 'Clone to Template…', and then follow the wizard.
