Installation Procedure

The NMX-M installation process is now simplified using a one-click shell script that ensures full air-gapped support. The below are detailed instructions on how to seamlessly install NMX-M on your dedicated machines.

The NMX-M package is distributed as a single tarball: NMX-M-<VERSION>.tar.gz.

Within this tarball, there are 2 tar archives: the VM image tar, and another tarball containing the NMX-M installation package. Untar the tar ball using tar -xvf NMX-M-<VERSION>.tar.gz.

To install the VM image, follow the steps below:

1. Extract NMX-MGR-<VERSION>-KVM.tar.gz

2. Mount the VM image ( .qcow2 file) using the appropriate KVM commands

   Example: Copy the Base image to - /var/lib/libvirt/images/

   Copy

   Copied!

               
   
               
   virt-install --connect qemu:///system --ram 512000 -n ubuntu1 --os-type linux --os-variant generic --vcpus=48 --disk path=/var/lib/libvirt/images/ubuntu-24.04-nmx-base.qcow2,format=qcow2,device=disk,bus=virtio --vnc --noautoconsole --import --network=bridge=br0,model=virtio,mac=<MAC> --check all=off
           

3. Log in to the nvidia user account on each of the three servers using the default password Nmx@123

4. Upon first login, you will be prompted to set a new password.

   Use the same new password on all three servers.

   Note: Changing the password will automatically log you out.

5. Re-login as the nvidia user using the new password on only one of the servers.

   The remaining installation steps will be performed only on this server.

To install the NMX Manager (NMX-M) software package, follow the steps below:

1. Copy the `NMX-MGR-M-<VERSION>.tar.gz ` file to one of the VMs. Once the tarball is in place, execute the following command to extract its contents:

   Copy

   Copied!

               
   
               
   tar -xzf NMX-MGR-M-*.tar.gz
           

2. Navigate to the created `NMX-M` directory:

   Copy

   Copied!

               
   
               
   cd NMX-M
           

3. Ensure you are in sudo mode before proceeding with the installation.

   Copy

   Copied!

               
   
               
   sudo su
           

4. Run the installation script:

   Copy

   Copied!

               
   
               
   ./install.sh
           

5. Install the required tools. The installation script will begin by installing all the necessary tools to support the installation. This includes:

   • Ansible

   • kubectl

   • zarf

   • helm

6. Configure the Kubernetes Cluster. You will be required to specify the IP addresses of three server Virtual Machines (VMs). Please note that exactly three VMs are required for this configuration. server1should be the IP address of the VM where the installation is currently being executed.

7. Additionally, during the setup process, you will be required to enter the SSH password.

8. Install RKE2. The script will proceed by running the RKE2 (Rancher Kubernetes Engine 2) installation.

9. Client Certificates for mTLS

   1. During installation, mTLS certificates are automatically created and can be found at this path: /opt/nvidia/nmx/Installation/Ansible/

   2. Follow NVIDIA NVOS User Manual for NVLink Switches → NVLink Switching → Cluster Management → Cluster Manager → Cluster Manager Commands to apply the certificates to both NMX-C and NMX-T.

10. Set the API user passwords. You will be asked twice to set passwords for the API users `ro-user ` and `rw-user`. Each password must be at least 8 characters long and include a mix of letters and numbers.

11. The script installs the infrastructure components.

12. The script installs the NMX-M Microservices.