HAMi 发布 v2.4.0 版本：异构计算虚拟化迈向新高度

apiVersion: v1
kind: Pod
metadata:
  name: gpu-pod
spec:
  containers:
    - name: ubuntu-container
      image: ascendhub.huawei.com/public-ascendhub/ascend-mindspore:23.0.RC3-centos7
      command: ["bash", "-c", "sleep 86400"]
      resources:
        limits:
          huawei.com/Ascend910B: 1 # requesting 1 Ascend
          huawei.com/Ascend910B-memory: 2000 # requesting 2000m device memor

apiVersion: v1
kind: Pod
metadata:
  name: gpu-pod
spec:
  containers:
    - name: ubuntu-container
      image: ascendhub.huawei.com/public-ascendhub/ascend-mindspore:23.0.RC3-centos7
      command: ["bash", "-c", "sleep 86400"]
      resources:
        limits:
          huawei.com/Ascend310P: 1 # requesting 1 Ascend
          huawei.com/Ascend310P-memory: 1024 # requesting 1024m device memory

{
        "nodeconfig": [
            {
                "name": "m5-cloudinfra-online02", //节点名称
                "devicememoryscaling": 1.8,
                "devicesplitcount": 10,
                "migstrategy":"none",
                "filterdevices": {
                  "uuid": ["GPU-54ea13f7-27d8-c1e4-4b86-9411e1283dff"],
                  "index": []
                }
            }
        ]
    }

(base) ➜  scheduler git:(master) ✗ kubectl describe  po nginx-deployment-7bd89f854c-5p2xv
Name:             nginx-deployment-7bd89f854c-5p2xv
Namespace:        default
Priority:         0
Service Account:  default
...
...
...
Events:
  Type    Reason            Age   From            Message
  ----    ------            ----  ----            -------
  Normal  Scheduled         6s    hami-scheduler  Successfully assigned default/nginx-deployment-67f97f4d9d-wtzzd to controller-node-1
  Normal  FilteringSucceed  7s    hami-scheduler  Successfully filtered to following nodes: [controller-node-1] for default/nginx-deployment-67f97f4d9d-wtzzd
  Normal  BindingSucceed    7s    hami-scheduler  Successfully binding node [controller-node-1] to default/nginx-deployment-67f97f4d9d-wtzzd
  Normal  Pulling           6s    kubelet         Pulling image "docker.m.daocloud.io/nginx:latest"
  Normal  Pulled            5s    kubelet         Successfully pulled image "docker.m.daocloud.io/nginx:latest" in 1.075344559s (1.075360343s including waiting)
  Normal  Created           5s    kubelet         Created container nginx
  Normal  Started           5s    kubelet         Started container nginx

(base) ➜  scheduler git:(master) ✗ kubectl describe po nginx-deployment-8dc9c4774-txbzf
Name:             nginx-deployment-8dc9c4774-txbzf
Namespace:        default
Priority:         0
...
...
...
Events:
  Type     Reason            Age   From            Message
  ----     ------            ----  ----            -------
  Warning  FailedScheduling  8s    hami-scheduler  0/2 nodes are available: 1 node(s) had untolerated taint {node.kubernetes.io/unreachable: }. preemption: 0/2 nodes are available: 1 No preemption victims found for incoming pod, 1 Preemption is not helpful for scheduling.
  Warning  FilteringFailed   9s    hami-scheduler  no available node, all node scores do not meet