Managing Lite Server AI Plug-ins

Scenario

NodeTaskHub is a plug-in that manages elastic nodes efficiently. It sends tasks and enables automated O&M for ModelArts Lite Server nodes. It supports high-frequency operations such as Ascend software upgrade, real-time detection, and fault diagnosis, reducing manual intervention risks and ensuring stable and efficient AI service processes.

The Lite Server task center provides you with multiple task templates for creating tasks. Task delivery depends on the NodeTaskHub plug-in installed on the Lite Server node. For some public images of the Lite Server, the NodeTaskHub plug-in is preset. You can choose to automatically install the plug-in when purchasing the Lite Server. If it is not installed, manually install it by referring to the following content.

Constraints

Currently, only Ascend Snt9b and Ascend Snt9b23 nodes are supported.
The node must be in the running state.
The Docker service is required for the plug-in. Ensure that the Docker environment has been installed on the node. The Docker environment has been installed in the public OS image of Lite Server. However, if it is not installed in your custom image, perform operations in Physical Machine Environment Configuration.
The plug-in occupies port 25317 of the node.

Procedure

Log in to the ModelArts console.
In the navigation pane on the left, choose Lite Servers under Resource Management. On the displayed page, click the Agents tab.
Figure 1 Agent Management
Locate a running node and view the plug-in installation status. If the plug-in is not installed, click Install in the Operation column.
Select the plug-in version as prompted and copy the one-click installation command to the node to install the NodeTaskHub.
After the plug-in is installed, return to the Agents tab. The plug-in is in the Running state.
Figure 2 Running plug-in