Registering an Edge Node

Constraints

• NPUs on the same node support only the same partition specification.
• You cannot perform virtualization partition on the NPU of an IEF-managed edge node by simply upgrading the NPU plug-in to v2.1.0. Uninstall and reinstall the node, manually partition the NPU, and then manage the node on IEF again.
• The NPU driver version must be later than 22.0. You can go to the driver path (for example, /usr/local/Ascend/driver) and run the cat version.info command to view your driver version.
• If you enable the NPU sharing mode, you need to set the container sharing mode for all chips on the NPU.
• NPU virtualization partition is not supported if the D310P-share is used.
• The healthy chips displayed refer to physical chips. If NPU virtualization partition or sharing is used, the number of chips after virtualization partition or sharing is displayed.

Registering an Edge Node

1. Log in to the IEF console, and click Switch Instance on the Dashboard page to select a platinum service instance.
2. In the navigation pane, choose Managed Resources > Edge Nodes. Then, click Register Edge Node in the upper right corner.
3. Configure basic edge node information.
   Figure 1 Basic edge node information (1)
   
   • Node Name: name of an edge node. The name can contain 1 to 64 characters, including letters, digits, hyphens (-), and underscores (_).
   • Description: Optional. Enter the description of the edge node.
   • Tags: Tags can be used to classify resources, facilitating resource management.
   • AI Accelerator Card

     Ascend AI accelerator card: supports edge nodes that support Ascend processors and NPU specifications of Ascend 310P and Ascend 310P virtualization partition. If you need to use Ascend 310, 310B, 910, 310P, 310P-share, or 310P-1c, first select the AI accelerator card and then select the NPU type.

     AI accelerator cards support the NPU types listed in the following table:

     Table 1 NPU types

     Type	Description
     Ascend 310	Ascend 310 chips
     Ascend 310B	Ascend 310B chips
     Ascend 910	Ascend 910 chips
     Ascend 310P	Ascend 310P chips
     Ascend 310P-share	Ascend 310P chips that support multi-container sharing
     Ascend 310P-1c	Ascend 310P chips after virtualization partition based on the template vir01
     Ascend 310P-2c	Ascend 310P chips after virtualization partition based on the template vir02
     Ascend 310P-2c.1cpu	Ascend 310P chips after virtualization partition based on the template vir02_1c
     Ascend 310P-4c	Ascend 310P chips after virtualization partition based on the template vir04
     Ascend 310P-4c.3cpu	Ascend 310P chips after virtualization partition based on the template vir04_3c

     NVIDIA GPU: Select this option if your edge node is equipped with an NVIDIA GPU.

     Not installed: Select this option if your edge node does not use any AI accelerator card.

     

     If you select NVIDIA GPU but your edge node is not equipped with any NVIDIA GPU, the edge node cannot be managed by IEF.

     If your edge node has been equipped with a GPU, install and configure the GPU driver on the edge node before you can manage it on IEF. For details, see Installing and Configuring a GPU Driver.

   • Bind Device: Bind an end device to the edge node. If you have not registered an end device, register one by referring to End Device Management. If you do not bind an end device in this step, you can bind one after the edge node is registered.
   • Docker: If you enable Docker, containerized applications can be deployed.
   • Listening Address

     The address on which the built-in MQTT broker of the edge node listens. It is used for edge-cloud messaging. For details about edge-cloud messaging, see Device Twin Working Principles and Edge-Cloud Message Overview.

     By default, local NICs lo (localhost) and docker0 are listened to. You can specify an NIC name or IP address to set the NIC to be listened to. Multiple NICs or IP addresses can be added.

   Figure 2 Basic edge node information (2)
   

   Currently, system and application logs can be collected from edge nodes.

   • System Logs: logs generated by the IEF software (such as edge-core, edge-logger, and edge-monitor) installed on edge nodes.
   • Application Logs: logs generated by applications deployed on edge nodes.

   Set the following parameters for both system and application logs:

   • Maximum Log Size (MB): maximum size of a log file, in MB. The default value is 50. The value range is 10–1000. If a log file reaches the specified size, the system compresses the log file and dumps it to a specified directory on the host.
     • System logs are stored in the /var/IEF/sys/log/ directory on edge nodes and will be dumped to AOM when the maximum log size is reached.
     • Application logs include the standard output of containers and the logs mounted to /var/IEF/app/log on the edge node. These logs will be dumped to AOM when the maximum log size is reached.
   • Log Rotation: log dump interval. You can select Daily, Weekly, Monthly, or Yearly. Log files will be dumped when either Maximum Log Size (MB) or Log Rotation is reached.
   • Logs Saved: maximum number of logs being retained at a specified period. The default value is 5. The value range is 1–10. Once this number is reached, when new log files are dumped in, old log files will be deleted on a first-in first-out (FIFO) basis.
   • Save Logs to AOM

     You can enable or disable log uploading to AOM. If it is enabled, you can view the logs in AOM. For details, see Viewing Logs on AOM.

4. Select I have read and agree to the Huawei Cloud Service Level Agreement. Click Register in the lower right corner.
   Registration using a certificate: Download the configuration file and EdgeCore Installer, which will be used in Managing an Edge Node.

   Figure 3 Downloading the configuration file and EdgeCore Installer
   

   1. Click Download Edge node name.tar.gz.
   2. Select the CPU architecture of your edge node, and click Download EdgeCore Installer.

5. In the lower right corner, select I've finished downloading and click Finish.

   The edge node is in the Unconnected state because the EdgeCore Installer downloaded in Registering an Edge Node has not been installed. For details, see Managing an Edge Node.

   Figure 4 Unconnected edge node
   

Follow-up Operation

After the registration is complete, you need to manage the edge node. For details, see Managing an Edge Node.