Part 1: Preparing server environment#

This part describes the process of preparing the server environment for successful platform installation.

Note

If Ubuntu 20.04 is installed while NVIDIA GPU is plugged in, the latest NVIDIA GPU driver will be installed automatically.

If you plug in the NVIDIA GPU to your system after OS installation you must install the NVIDIA GPU driver manually. Installation instructions are described below.

Make sure your server is exclusively used for the Intel® Geti™ platform.

On the server, open a terminal and execute the following steps.

Step 1: Ensure using a sudo account to install the Intel® Geti™ platform#

The user’s account used to install the platform requires administrative privileges and must be included in the sudoers lists. Ensure it is present in /etc/sudoers.

If not, create the new account:

adduser <username>

or

sudo adduser <username>

When host OS is Ubuntu, please add it to group sudo:

usermod -aG sudo <username> or sudo usermod -aG sudo <username>

When host OS is Red Hat, please add it to group wheel:

usermod -aG wheel <username> or sudo usermod -aG wheel <username>

Then log out and log in again with your newly created account.

Check sudo by clearing the sudo cache:

sudo -k

and by running any command with sudo, e.g.:

sudo ls

This command should simply work.

Ensure that /dev/mem is accessible by the root user.

You can check it by running the following command and observing if it returns permission errors:

sudo head /dev/mem

Step 2: (Optional) Configure proxy#

Note

Complete this step only if your Intel® Geti™ server will access the internet from behind a proxy.

One way to configure a proxy is to create a file /etc/environment, e.g. with vi:

sudo vi /etc/environment

With all proxies that are applicable in your case. Sample /etc/environment file:

http_proxy=http://USERNAME:PASSWORD@SERVER:PORT/
https_proxy=http://USERNAME:PASSWORD@SERVER:PORT/
no_proxy=localhost,127.0.0.1,<Intel® Geti™ server IP address>
HTTP_PROXY=http://USERNAME:PASSWORD@SERVER:PORT/
HTTPS_PROXY=http://USERNAME:PASSWORD@SERVER:PORT/
NO_PROXY=localhost,127.0.0.1,<Intel® Geti™ server IP address>

Confirm proxy settings:

source /etc/environment
env | grep -i proxy

When host OS is Ubuntu please create file /etc/apt/apt.conf and add content:

sudo vi /etc/apt/apt.conf
# Press i
# Copy and press [right-click] below content and edit as necessary
Acquire::http::proxy "http://USERNAME:PASSWORD@SERVER:PORT/";
Acquire::https::proxy "http://USERNAME:PASSWORD@SERVER:PORT/";
# Press [Escape]
# Press [SHIFT + ZZ]

When host OS is Red Hat please update file /etc/dnf/dnf.conf and add content:

sudo vi /etc/dnf/dnf.conf
# Press i
# Copy and press [right-click] below content and edit as necessary at the end of file
proxy="http://USERNAME:PASSWORD@SERVER:PORT/";
# Press [Escape]
# Press [SHIFT + ZZ]

Step 3: Create a folder where all Intel® Geti™ data will be stored#

To create a folder, run the following command:

mkdir /<path>/data

Note

<path> is a mounted drive or directory with read/write access permission and adequate storage space for the dataset that you are planning to use, e.g., run df -h.

Note

While using the Intel® Geti™ platform after the installation, if you come close to running out of space (you will receive a notification), use this command to remove some temporary files: rm -rf /data-xfs/ie_storage/imports/*

To give folder required permissions, run the following command:

chmod 750 /<path>/data

Step 4: Install GPU drivers - NVIDIA#

Check NVIDIA GPU driver.
```
nvidia-smi
```
If nvidia-smi is not installed on the system, install NVIDIA GPU driver. The recommended version is 525 (nvidia-driver-525).

Run the commands below in case of Ubuntu:
```
sudo apt install nvidia-dkms-525
sudo apt install nvidia-driver-525
```
or for Red Hat:
```
sudo dnf install nvidia-dkms-525
sudo dnf install nvidia-driver-525
```
Reboot the system after installing the drivers so that they load into the kernel.
```
sudo reboot
```

Set up NVIDIA Container Toolkit with a stable repository and GPG key.

In case of Ubuntu:

# Check if NVIDIA gpgkey is added to the apt-key repository
apt-key list
# Look for NVIDIA Corporation entry
# If NVIDIA is not listed, run:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -

In case of Red Hat:

curl -s -L https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo | \
sudo tee /etc/yum.repos.d/nvidia-container-toolkit.repo

Step 5: Install GPU drivers - Intel#

Check Intel GPU driver.
```
xpu-smi
```
If xpu-smi is not installed on the system, install the latest Intel GPU driver as described in the Intel Data Center FPU installation manual.
Reboot the system after installing the drivers so that they load into the kernel.
```
sudo reboot
```

Note

Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.

Step 6: Configure Firewall#

For successful installation and operation of K3s, adjusting the firewall settings on the host machine is necessary. Follow the instructions specific to your host’s operating system:

Red Hat-based systems (RHEL): Please consult the K3s installation requirements for RHEL here. The guide includes steps to properly disable or configure the firewall to allow K3s to function correctly.
Debian-based systems (Ubuntu): For hosts running Ubuntu, follow the K3s installation requirements detailed here to disable or adjust the firewall accordingly.