Upgrading the Ascend Driver and Firmware Version on Lite Server

Scenario

This section describes how to deliver an Ascend driver and firmware upgrade task on the Lite Server task center. In this way, you can upgrade the driver and firmware on the Snt9b server in one-click mode.

Constraints

• Currently, only Ascend Snt9b and Ascend Snt9b23 nodes are supported.
• The upgrade may cause service interruption. Ensure that there is no running service on the node before the upgrade. After the upgrade, restart the node.
• The MCU version on the node will be upgraded accordingly during the upgrade.
• If the official driver and firmware versions are used on the node and the upgrade fails, they can be rolled back to the original versions. If the driver and firmware on the node are damaged or the official versions are not used, the query will fail. In this case, the upgrade task can still be delivered. However, the driver and firmware cannot be rolled back once the upgrade fails, and you need to contact Huawei O&M engineers.
• The driver and firmware are compatible with the Ascend software package, such as CANN and MindSpore. Ensure that the upgraded versions are compatible with the Ascend software package used in the service. For details, see Table 1.

  Table 1 Component compatibility

  CANN Version	Ascend HDK Version
  CANN 8.0.RC3	Ascend HDK 24.1.RC3 Ascend HDK 24.1.RC2 Ascend HDK 24.1.RC1 Ascend HDK 23.0.0/23.0.X
  CANN 8.0.0	Ascend HDK 24.1.0 Ascend HDK 24.1.RC3 Ascend HDK 24.1.RC2 Ascend HDK 24.1.RC1 Ascend HDK 23.0.0/23.0.X
  CANN 8.1.RC1	Ascend HDK 25.0.RC1 Ascend HDK 24.1.0 Ascend HDK 24.1.RC3 Ascend HDK 24.1.RC2 Ascend HDK 24.1.RC1 Ascend HDK 23.0.X
  CANN 8.2.RC1	Ascend HDK 25.2.0 Ascend HDK 25.0.RC1 Ascend HDK 24.1.0 Ascend HDK 24.1.RC3 Ascend HDK 24.1.RC2

Prerequisites

This operation depends on the Lite Server AI plug-in pre-installed on the node. Install the plug-in by referring to Managing Lite Server AI Plug-ins.

Procedure

1. Log in to the ModelArts console.
2. In the navigation pane on the left, choose Lite Servers under Resource Management. On the displayed page, click the Task Center tab.
   Figure 1 Task center
   

3. Click Create Task in the upper left corner. On the displayed Job Templates page, locate Ascend Software Upgrade, and click Create Task.
   Figure 2 Task template
   

4. On the displayed page, set Name, Description, Server Model, and Type. Select a node, and click OK in the displayed dialog box. The driver firmware version query task will be delivered to the selected node, which takes about 1 minute. Then, you can obtain the actual driver firmware information.

   Table 2 Parameters for creating a task

   Parameter	Description
   Name	The system automatically enters the name of the pressure test task. You can customize the task name.
   Description	Enter the task description for quick search.
   Server Model	Only Ascend Snt9b and Ascend Snt9b23 are supported.
   Type	You can select Single node or Integrated rack, or search by keyword. Select the target node to be upgraded in the node list (batch selection supported) and click OK. A query task will be delivered to the selected node to query the driver and firmware versions and CANN information. Wait for the query result, which takes about one minute.
   Software Version	Select the target driver and firmware versions from the drop-down list. Check the component compatibility by referring to Table 1 to prevent service interruption caused by upgrade failure. You can also deliver a fault diagnosis task for Ascend devices by referring to Lite Server Node Fault Diagnosis. The compatibility between the driver & firmware and CANN will be automatically diagnosed.

5. Select the version to be upgraded from the drop-down list, click Next, confirm the upgrade information, select automatic or manual restart upon upgrade, and click OK to deliver the upgrade task. It takes about 10 minutes to complete the upgrade after the task is delivered.
6. During the upgrade, you can view the task execution status in the Task Center tab. Click the task name to access its details page, where you can view the task details and logs.
7. Restart is required upon upgrade. If manual restart is selected, run the reboot command on the node. The process takes about 10 minutes.
8. Run the command on the node to check whether the driver is loaded. If the command below is returned, the load is successful. Otherwise, contact Huawei technical support.

   npu-smi info

   Figure 3 Checking whether the driver is loaded