Task Pending While Allocating an Eight-Card Job When the Number of Chips Corresponding to Allocatable.huawei.com/Ascend910 in the Node Information Is 8

Symptom

Run the kubectl describe node {node name} command to view the node information. The number of chips corresponding to Allocatable.huawei.com/Ascend910 is 8. An Eight-Card Job is allocated, and the task is in pending status.

Capacity:
  cpu:                   72
  ephemeral-storage:     1843598940Ki
  huawei.com/Ascend910:  8
  hugepages-1Gi:         0
  hugepages-2Mi:         0
  memory:                659447564Ki
  pods:                  110
Allocatable:
  cpu:                   72
  ephemeral-storage:     1699060780291
  huawei.com/Ascend910:  8
  hugepages-1Gi:         0
  hugepages-2Mi:         0
  memory:                659345164Ki
  pods:                  110

Cause Analysis

Common faults that are not detected by Ascend Device Plugin may exist on the node.

Solution

Rectify the fault by referring to Obtaining Information About Available Devices in a Cluster.