Overview

ModelArts offers several plug-ins to help you expand resource pool functions as needed.

Default Plug-ins

The plug-ins are installed by default when you create a dedicated resource pool.

Plug-ins installed by default in a resource pool cannot be uninstalled.

**Table 1** Default plug-ins
Plug-in	Description
Node Fault Detection (ModelArts Node Agent)	ModelArts Node Agent is a plug-in for monitoring cluster node exceptions, also, a component for connecting to third-party monitoring platforms. It is a daemon that runs on each node to collect node problems from different daemon processes.
AI Suite (ModelArts Device Plugin)	The CCE AI suite, Ascend NPU, is a device management plug-in that supports Huawei NPUs in containers. When you enable Lite Cluster resources, this plug-in is automatically downloaded only when the instance specification type is set to Ascend.
Volcano Scheduler	Volcano is a batch scheduling platform based on Kubernetes. It provides a series of features required by machine learning, deep learning, bioinformatics, genomics, and other big data applications, as a powerful supplement to Kubernetes capabilities.

Installing the Plug-in Manually

You can install plug-ins to extend resource pool functions as required.

**Table 2** Default plug-ins
Plug-in	Description
Cluster Autoscaler	Cluster Autoscaler is a plug-in for elastic scaling of ModelArts resource pools in a cluster. It can be used to scale in or out node pools based on user-defined rules.

Plug-in Lifecycle

Status	Status Attribute	Description
Installing	Intermediate	The plug-in is being deployed. If all instances cannot be scheduled due to incorrect plug-in configuration or insufficient resources, the system sets the plug-in status to Unavailable 10 minutes later.
Running	Stable	The plug-in is running, all plug-in instances are deployed, and the plug-in can be used properly.
Upgrading	Intermediate	The plug-in is being upgraded.
Unavailable	Stable	The plug-in is abnormal and cannot be used. You can click the status to view the failure cause.
Deleting	Intermediate	The plug-in is being deleted. If this state stays for a long time, an exception occurred.

Searching for a Plug-in on the Plug-in Square

The ModelArts Plug-in Square provides various plug-ins. You can view the plug-in details and install them to a specified resource pool as needed.

**Table 3** Supported operations on the Plug-in Square
Operation	Description	Procedure
Searching for and viewing a plug-in	Search for and view a plug-in.	Log in to the ModelArts console. In the navigation pane on the left, choose Add-ons. Choose a resource type from the drop-down list to filter plug-ins, or enter a keyword in the search box to search for a plug-in.
Viewing plug-in details	View the plug-in details, including the plug-in introduction and component list.	Log in to the ModelArts console. In the navigation pane on the left, choose Add-ons. Click the plug-in name to view its details.
Installing a plug-in	Certain plug-ins can be manually installed.	Log in to the ModelArts console. In the navigation pane on the left, choose Add-ons. Locate the target plug-in and click Install. In the displayed dialog box, select the resource type of the plug-in to be installed. For some plug-ins, you also need to select a plug-in version. Set the information and click Next. Dedicated cluster: Install the plug-in to a resource pool. The supported resource pool types vary depending on the plug-in. See the supported types on the GUI accordingly. Dedicated node: Install the plug-in to a specific node in the resource pool. Perform operations and run commands as prompted. Configure related parameters. The configurations vary depending on the plug-in. For details, see section"Plug-ins".

Viewing the Lite Cluster Plug-in on the Resource Pool Details Page

In the Plug-ins tab of the resource pool details page, perform the operations described in Table 4.

**Table 4** Related operations
Operation	Description	Procedure
Querying the plug-ins	View all plug-ins of a resource pool. On this page, you can view plug-in details, install, upgrade, and uninstall plug-ins, and manage plug-ins in a centralized manner.	Log in to the ModelArts console. In the navigation pane on the left, choose Lite Cluster under Resource Management. Click the resource pool name to access its details page. In the navigation pane on the left, choose Plug-ins.
Viewing plug-in details	View the plug-in details, including the plug-in introduction and component list.	Log in to the ModelArts console. In the navigation pane on the left, choose Lite Cluster under Resource Management. Click the resource pool name to access its details page. In the navigation pane on the left, choose Plug-ins. Click the plug-in name to view its details.
Plug-ins installed by default	When you create a resource pool, certain plug-ins are installed by default.	Enabling Lite Cluster Resources
Installing the plug-in manually	Install the specified plug-in in the resource pool.	Method 1: Install the plug-in when you enable Lite Cluster resources. For details, see Enabling Lite Cluster Resources. Method 2: Log in to the ModelArts console. In the navigation pane on the left, choose Lite Cluster under Resource Management. Click the resource pool name to access its details page. In the navigation pane on the left, choose Plug-ins. Locate the plug-in to be installed and click Install, as shown in Figure 1. In the displayed dialog box, configure the parameters. Currently, Lite Cluster allows you to manually install the elastic cluster engine plug-in. For details about the parameters, see Table 1.
Editing a plug-in	Edit plug-in parameters.	Log in to the ModelArts console. In the navigation pane on the left, choose Lite Cluster under Resource Management. Click the resource pool to access its details page. In the navigation pane on the left, choose Plug-ins. Locate the plug-in to be edited in the list and click Edit. The configurations vary depending on plug-ins. For details, see "Plug-ins". Only the following plug-in versions can be edited: ModelArts Node Agent 7.2.0 or later AI suite (Ascend NPU) 2.1.53 or later Volcano Scheduler 1.17.11 or later Cluster Autoscaler 0.1.13 or later Click OK.
Upgrading a plug-in	Upgrade the plug-in to the latest version.	Log in to the ModelArts console. In the navigation pane on the left, choose Lite Cluster under Resource Management. Click the resource pool to access its details page. In the navigation pane on the left, choose Plug-ins. Locate the plug-in to be upgraded in the list and click Upgrade. Currently, Lite Cluster allows you to manually install the elastic cluster engine plug-in. For details about the parameters, see Table 1. Click OK. CAUTION: Plug-ins are deployed based on Helm templates. To modify or upgrade plug-ins, you need to use the plug-in list on the ModelArts console or the open plug-in management APIs. Do not manually modify related resources on the CCE server. Otherwise, exceptions may occur, for example, parameter settings may be lost or overwritten after the upgrade. During the plug-in upgrade, some functions of the resource pool may be affected. You should check the status and version compatibility of all external dependencies before the upgrade and reserve enough time for the upgrade. For details about the impact, see the plug-in description.
Uninstalling a plug-in	Uninstall a plug-in from the resource pool. This operation cannot be undone.	Log in to the ModelArts console. In the navigation pane on the left, choose Lite Cluster under Resource Management. Click the resource pool to access its details page. In the navigation pane on the left, choose Plug-ins. Locate the plug-in to be uninstalled in the list and click Uninstall. In the displayed dialog box, enter DELETE and click OK.

Figure 1 Installing a plug-in

FAQ

If the plug-in must be installed is unavailable or is being installed or deleted for a long time, you can click the resource pool name to view the basic information. In the CCE cluster area, click the CCE cluster in the resource pool.
Go to the plug-in center, locate the target plug-in, and click it to view the details. In the instance list, click the abnormal instance and check the exception cause.
If an optional plug-in is unavailable or is being installed or deleted for a long time, uninstall the plug-in and reinstall it. If the plug-in is still unavailable after the re-installation, view the exception details by referring to the previous step.
If the fault persists, contact ModelArts technical personnel.

Parent topic: Lite Cluster Plug-in Management

Previous topic: Lite Cluster Plug-in Management

Next topic: Node Fault Detection (ModelArts Node Agent)