默认 kube-scheduler 用的是 LeastRequested：空闲越多越优先，我们需要改成：MostAllocated（已用越多、得分越高、越优先）。

我们目前想根据线上环境调整下策略，实际需求就是使用GPU的POD优先调度到没使用完GPU的节点。

需求讲完了，我们来做下实际配置，其实配置比较简单，就几个简单步骤。

1.创建调度策略配置文件

tee /etc/kubernetes/scheduler-config.yaml<<EOF
apiVersion: kubescheduler.config.k8s.io/v1
kind: KubeSchedulerConfiguration
clientConnection:
  kubeconfig: /etc/kubernetes/scheduler.conf
profiles:
- schedulerName: default-scheduler
  plugins:
    score:
      enabled:
      - name: NodeResourcesFit
        weight: 100
      disabled:
      - name: LeastRequestedPriority
      - name: BalancedAllocation
  pluginConfig:
  - name: NodeResourcesFit
    args:
      scoringStrategy:
        type: MostAllocated
        resources:
        - name: nvidia.com/gpu
          weight: 100
EOF

由于我们只关注GPU资源，所以我们把权重全部给了GPU，资源名字需要和你自己的环境中的GPU资源名字保持一致。

2.修改调度器配置

cat /etc/kubernetes/manifests/kube-scheduler.yaml
apiVersion: v1
kind: Pod
metadata:
  creationTimestamp: null
  labels:
    component: kube-scheduler
    tier: control-plane
  name: kube-scheduler
  namespace: kube-system
spec:
  containers:
  - command:
    - kube-scheduler
    - --config=/etc/kubernetes/scheduler-config.yaml # 添加此处配置
    - --authentication-kubeconfig=/etc/kubernetes/scheduler.conf
    - --authorization-kubeconfig=/etc/kubernetes/scheduler.conf
    - --bind-address=127.0.0.1
    - --kubeconfig=/etc/kubernetes/scheduler.conf
    - --leader-elect=true
    image: registry.k8s.io/kube-scheduler:v1.33.0
    imagePullPolicy: IfNotPresent
    livenessProbe:
      failureThreshold: 8
      httpGet:
        host: 127.0.0.1
        path: /livez
        port: 10259
        scheme: HTTPS
      initialDelaySeconds: 10
      periodSeconds: 10
      timeoutSeconds: 15
    name: kube-scheduler
    readinessProbe:
      failureThreshold: 3
      httpGet:
        host: 127.0.0.1
        path: /readyz
        port: 10259
        scheme: HTTPS
      periodSeconds: 1
      timeoutSeconds: 15
    resources:
      requests:
        cpu: 100m
    startupProbe:
      failureThreshold: 24
      httpGet:
        host: 127.0.0.1
        path: /livez
        port: 10259
        scheme: HTTPS
      initialDelaySeconds: 10
      periodSeconds: 10
      timeoutSeconds: 15
    volumeMounts:
    - mountPath: /etc/kubernetes/scheduler.conf
      name: kubeconfig
      readOnly: true    # 挂在配置到POD内部
    - mountPath: /etc/kubernetes/scheduler-config.yaml
      name: scheduler-config
      readOnly: true
  hostNetwork: true
  priority: 2000001000
  priorityClassName: system-node-critical
  securityContext:
    seccompProfile:
      type: RuntimeDefault
  volumes:
  - hostPath:
      path: /etc/kubernetes/scheduler.conf
      type: FileOrCreate
    name: kubeconfig#挂配置到POD内
  - hostPath:
      path: /etc/kubernetes/scheduler-config.yaml
      type: File
    name: scheduler-config
status: {}

上述配置修改好以后scheduler的POD就会自动重启，接着我们就可以进行POD调度测试了。