Skip to main content

Prepare the Jetson Host

This page describes how to prepare an NVIDIA Jetson device so that it can register with Palette as an Edge host in agent mode.

Prerequisites​

  • An NVIDIA Jetson device that meets the Jetson Requirements. This guide is validated on the NVIDIA Jetson AGX Thor Developer Kit.

  • Console access to the device, such as a monitor and keyboard. You use the console to complete the operating system setup and to reconnect if a network change drops your SSH session.

  • A USB drive for the JetPack installer image.

Prepare the Host​

The NVIDIA Jetson AGX Thor Developer Kit runs NVIDIA JetPack, which provides a Jetson Linux (L4T) operating system built on Ubuntu. The FIPS-compliant version of agent mode is available only for Red Hat Enterprise Linux and Rocky Linux 8. On a JetPack (Ubuntu) host, use the non-FIPS agent.

  1. Install JetPack by following the NVIDIA Jetson AGX Thor Developer Kit Quick Start Guide. Download the installer image from the JetPack download page, write it to a USB drive, and then select Install on NVMe to install the operating system on the NVMe SSD.

  2. Complete the operating system setup. During setup, the installer prompts you to enable Ubuntu Pro. Palette does not require Ubuntu Pro. On a Jetson device, only the esm-infra and esm-apps services apply. Do not enable the FIPS kernel, FIPS updates, or the real-time kernel, because those services try to replace the Jetson Linux (L4T) kernel.

  3. Install the following packages, which the Palette agent requires. This list matches the prerequisites in Install Palette Agent.

    Because JetPack is Ubuntu-based, you can install the packages with apt.

    sudo apt-get update && \
    sudo apt-get install --yes --no-install-recommends \
      bash \
      jq \
      zstd \
      rsync \
      systemd-timesyncd \
      conntrack \
      iptables \
      rsyslog

  4. Enable the required systemd services.

    sudo systemctl enable --now systemd-timesyncd
    sudo systemctl enable --now systemd-resolved
    sudo systemctl enable --now rsyslog

    JetPack already runs systemd-networkd, so you do not need to enable it.

  5. (Optional) If you plan to use overlay networks, or you want Palette to manage DNS or static IP addresses, configure systemd-resolved and systemd-networkd. Refer to Configure networkd to Prepare Host for Overlay Network.

    warning

    On a Jetson device, NetworkManager manages the network interface by default. If you move the interface that your SSH session uses to systemd-networkd, the connection can drop. Before you make the change on a remote device, confirm that you have console access, or configure systemd-networkd with your network settings first. If your SSH session stops responding, press ENTER, type ~. to close it, and then reconnect from the console.

  6. Create a Palette registration token and keep it available for the registration step.

Validate​

Confirm the operating system versions on the device. The value that the device reports at first boot, such as 38.0.0-gcid-..., is the Unified Extensible Firmware Interface (UEFI) firmware version, not the operating system version.

  1. Check the authoritative Jetson Linux (L4T) release.

    cat /etc/nv_tegra_release
    Example Output
    # R39 (release), REVISION: 2.1, GCID: 46758480, BOARD: generic, EABI: aarch64, DATE: Fri Aug  7 05:54:22 AM UTC 2026
    # KERNEL_VARIANT: oot
    TARGET_USERSPACE_LIB_DIR=nvidia
    TARGET_USERSPACE_LIB_DIR_PATH=usr/lib/aarch64-linux-gnu/nvidia

    In this example, the release R39 and the revision 2.1 identify L4T r39.2.1.

  2. Check the Ubuntu base version.

    lsb_release --all
    Example Output
    No LSB modules are available.
    Distributor ID: Ubuntu
    Description:    Ubuntu 24.04.4 LTS
    Release:        24.04
    Codename:       noble

  3. If you installed the JetPack SDK meta-package, check the JetPack version.

    apt-cache show nvidia-jetpack

    On the base operating system, the package is not installed. Compare the L4T release with the validated versions, or look up the matching JetPack version in the JetPack Archive.

Next Steps​

The device is now ready to register with Palette and run an AI model. Continue to Register a Jetson Host and Serve a Model.