How to Create a Basic Lingjun Cluster - Intelligent Computing LINGJUN

A basic Lingjun cluster provides a clean computing environment where you can log on to nodes through SSH and set up your own development environment.

Prerequisites

You have completed Alibaba Cloud account registration and identity verification, and purchased Lingjun nodes. For more information, see Purchase Lingjun Nodes.

Create a Basic Lingjun Cluster

Step 1: Configure the cluster and group

Log on to the Intelligent Computing Lingjun console. In the left navigation bar, choose Resources and Nodes > Cluster Management, and click Create Cluster > Basic Lingjun Cluster Service.
Configure the following key parameters:
- HPNZone ID: Enter the cluster ID obtained when purchasing the edge zone.
- Resource group: Select the resource group that the cluster belongs to for categorization and management.
Click Create Group and configure the following key group info. Other parameters can be configured as needed.
- Default Model: Select the purchased node instance type from the drop-down list.
- Node Instance: Select purchased node resources to add to the current group.
- Logon Password: Set the root user logon password for cluster nodes to use in subsequent SSH connections. You can also use a key pair.

Click Save and Go to Next Step Network Configuration.

Step 2: Configure cluster network

For some regions and instance types, you must use Lingjun connection instances to connect Lingjun nodes to your Cloud Enterprise Network or virtual private cloud (VPC). Check the configuration page to determine whether Lingjun connection instances are required, and then follow the corresponding instructions.

No Lingjun connection instances required

VPC: Select an existing VPC. If no VPC is available, click Create VPC on the page.
Important
The VPC region must match the Lingjun node region.
vSwitch and Security group: Select the vSwitch created under the VPC. For more information, see Virtual Private Cloud and vSwitch and Use Security Groups.
Backend RDMA Network IP Version: Select the IP version for the backend RDMA network. Both IPv6 and IPv4 are supported. Some regions support only IPv4.

Configure Lingjun connection instances

Configure the Cluster CIDR Block.
- Cluster CIDR Block: A private CIDR block for internal cluster use, used to assign IP addresses to Lingjun nodes. Example: 172.16.0.0/12.
- Cluster Subnet: A subnet of the cluster CIDR block for more granular IP address assignment.
Configure Lingjun Connection.
Click Authorize Now to authorize the Lingjun service-linked role AliyunServiceRoleForEfloVcc, which allows Lingjun connection instances to access other cloud services.
- Instance ID: Select a purchased Lingjun connection instance from the drop-down list. If no Lingjun connection instance is available, click Buy Lingjun Connection Instance on the page.
  Important
  The region of the Lingjun connection instances must match the region of the Lingjun node.
- Cloud Enterprise Network (CEN): Select an existing CEN instance.
  Important
  Ensure that a transit router has been created in CEN and that the transit router region matches the Lingjun node region. For more information, see transit router instances.
Configure Network Information Monitoring.
- VPC: Select a VPC under CEN. You can add a VPC to CEN after creating it. Lingjun uses the vSwitch under this VPC to monitor the connectivity of Lingjun connection instances. Ensure that the vSwitch has at least one idle IP address.
  Important
  - The region of the VPC must match the region of the Lingjun node.
  - The CIDR blocks for the cluster network, internet segment, and monitoring network must not overlap. The monitoring network VPC must not conflict with other network environments connected to Lingjun, such as other VPCs or on-premises IDC CIDR blocks. For example, if the cluster CIDR block is 172.16.0.0/12, the VPC CIDR block can be 10.0.0.0/16.
  - Only VPCs added to a CEN can be selected from the drop-down list. For more information, see Create VPC Connection.
- vSwitch: Select the created vSwitch.

Click Save and proceed to the next step: Software Instance Basic Parameters.

Step 3: Configure Software Instance Basic Parameters

To use CPFS for Lingjun in hostPath mode, configure the File System ID and Mount Path as needed. Otherwise, keep the defaults and click Save and Go to Next Step: Confirm Configuration.

Step 4: Confirm Configuration and Create Cluster

After verifying that the cluster information, network configuration, and other parameters are correct, click Submit Configuration to create the cluster.

Cluster creation takes a few minutes. After the cluster is created, you are redirected to the cluster management page, where you can view the cluster status in the cluster list.

Connect to a Lingjun node remotely through SSH

Method 1: Log on via the Lingjun console

Log on to the Lingjun console. In the left navigation bar, choose Resources and Nodes > Node Management.
In the Actions column next to the target node ID, click Remote Logon.
- The logon username is root.
- The logon password is the password set during cluster creation.

Method 2: Connect Using an ECS Jump Server

Create an ECS instance in the same VPC as a jump server to connect to a Lingjun node through SSH.

Create an ECS instance in the same VPC as the Lingjun cluster.
For more information, see Purchase an ECS instance from the console. Note: Select the VPC used by the Lingjun cluster when selecting a VPC.
Configure security group rules for the ECS instance to allow SSH port 22.
Add the following inbound rules to the security group:
- Port range: SSH (22)
- Authorization object: 0.0.0.0/0 (in production environments, specify a specific IP range).
- Authorization policy: Allow
Connect to the ECS instance through SSH first.
For detailed steps, see Log on to a Linux instance using Workbench. For more connection methods, see Select an ECS remote connection method.
Connect to the Lingjun node from the ECS instance through SSH.
1. In the Lingjun console, choose Resources and Nodes > Node Management, click the target node name, and view the private IP in the Network Information area.
2. Run the following command in the ECS instance terminal to connect to the Lingjun node:
```
ssh root@<Lingjun node private IP>
```
3. Enter the logon password set during cluster creation to log on to the Lingjun node.

Method 3: Connect using an Internet NAT gateway

By configuring DNAT entries for an Internet NAT gateway, you can connect to a Lingjun node directly from the Internet through SSH.

Purchase and configure an Internet NAT gateway.
For more information, see Internet NAT Gateway.
Configure DNAT entries to implement port mapping.
- Select the Elastic IP Address associated with the Internet NAT gateway.
- In the Lingjun console, choose Resources and Nodes > Node Management, click the target node name, and view the private IP in the Network Information area.
- Select the private IP address of the Lingjun node by entering it manually.
- Customize the Internet port (for example, 2222), private network port (22), and protocol type (TCP).
Connect to the Lingjun node from a local terminal through SSH.
```
ssh root@<NAT Gateway EIP address> -p <Internet port>
```

Next Steps

After you connect to the Lingjun node, you can:

Install development environments and AI frameworks based on your requirements, such as PyTorch and TensorFlow.
Upload training data and model code to start AI model training. For example, read OSS files using the Alibaba Cloud SDK.