Add Machines

DGX Cloud Lepton provides a simple way to manage all your BYOC nodes in the node group page.

There are two ways to add your machines to DGX Cloud Lepton:

• Add machines via SSH access
• Add machines via local command

Prerequisites

• To bring your own compute to DGX Cloud Lepton, you need to have a node group created first. If you don't have one, refer to the node group document to request a node group.
• Only workspace admins can add machines to the node group.
• The machines should meet the requirements of BYOC.

Add Machines via SSH Access

The first way is to add your machines via SSH access. Follow the steps below:

1. Navigate to the detailed page of the node group you want to add machines to, click on the "Add Machines" button in the "Nodes" tab, and select "Add via SSH Access".

2. Obtain the IP address and username of the machine you want to add.

3. Input the IP address and username in the following format:

   username@ip_address
   username2@ip_address2
   

   You can add up to 10 machines at once by adding each machine on a new line.

4. Click on the "Import" button to start importing.

5. After importing the IP address and username, Lepton will generate a public key for each machine and return a command to add the machines to the node group.

6. Copy the command and run it on the machine to register the public key. Lepton will use this key to handle the rest of the process.

Add Machines via Local Command

You can also add your machines via local command, powered by GPUd:

1. Navigate to the detailed page of the node group you want to add machines to, click on the "Add Machines" button in the "Nodes" tab, and select "Add via Local Command".
2. Click on the "Generate Command" button in the pop-up window.
3. You will see the generated command like the following script and the expiration time:

   curl -fsSL https://pkg.gpud.dev/install.sh | sudo sh -s v0.5.0 && \
   sudo gpud up \
       --token ${TOKEN} \
       --endpoint ${ENDPOINT} \
       --node-group ${NODE_GROUP_ID}
   

4. Copy the command and run it on the machine you want to add. This command will install GPUd and use GPUd to join the node group.

Validate the Nodes

Lepton will automatically validate the nodes after adding them, showing status in the node list.

If the machine is added successfully, you can see the node status is "Healthy".

Release Machines

You can also release machines from the node group, which are represented as "Nodes" in the node group page.

1. Click on the menu button and select the "Release" option.
2. Select "Graceful Release" or "Release" option, input the reason, and submit the request.
3. The machines will be released from the node group, which will remove the dependencies installed during the BYOC process.