1
# 预选阶段 (Filtering) - 过滤不符合的节点
2
# - PodFitsResources: 检查资源是否足够
3
# - PodFitsHostPorts: 检查端口是否冲突
4
# - HostName: 检查节点名称
5
# - MatchNodeSelector: 检查节点选择器
6
# - NoDiskConflict: 检查存储冲突
7

8
# 优选阶段 (Scoring) - 选择最佳节点
9
# - LeastRequestedPriority: 优先选择资源少的节点
10
# - BalancedResourceAllocation: 平衡 CPU 和内存
11
# - ImageLocalityPriority: 优先选择镜像缓存的节点

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: nginx
5
spec:
6
  nodeSelector:
7
    disktype: ssd # 标签键值对
8
    region: us-east-1

1
# 给节点打标签
2
kubectl label nodes node-1 disktype=ssd
3
kubectl label nodes node-1 region=us-east-1
4

5
# 查看标签
6
kubectl get nodes --show-labels
7

8
# 删除标签
9
kubectl label nodes node-1 disktype-

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: nginx
5
spec:
6
  affinity:
7
    nodeAffinity:
8
      # 必须满足的条件
9
      requiredDuringSchedulingIgnoredDuringExecution:
10
        nodeSelectorTerms:
11
          - matchExpressions:
12
              - key: topology.kubernetes.io/zone
13
                operator: In
14
                values:
15
                  - us-east-1a
16
                  - us-east-1b
17

18
      # 优先满足的条件
19
      preferredDuringSchedulingIgnoredDuringExecution:
20
        - weight: 1
21
          preference:
22
            matchExpressions:
23
              - key: memory
24
                operator: Gt
25
                values:
26
                  - "8"

1
# 优先调度到有 redis 的节点
2
apiVersion: v1
3
kind: Pod
4
metadata:
5
  name: app
6
spec:
7
  affinity:
8
    podAffinity:
9
      preferredDuringSchedulingIgnoredDuringExecution:
10
        - weight: 100
11
          podAffinityTerm:
12
            labelSelector:
13
              matchLabels:
14
                app: redis
15
            topologyKey: kubernetes.io/hostname

1
# 分散 Pod，避免单点故障
2
apiVersion: v1
3
kind: Pod
4
metadata:
5
  name: nginx
6
spec:
7
  affinity:
8
    podAntiAffinity:
9
      requiredDuringSchedulingIgnoredDuringExecution:
10
        - labelSelector:
11
            matchLabels:
12
              app: nginx
13
          topologyKey: kubernetes.io/hostname

1
# Pod 拓扑分布约束
2
apiVersion: v1
3
kind: Pod
4
metadata:
5
  name: nginx
6
spec:
7
  topologySpreadConstraints:
8
    - maxSkew: 1
9
      topologyKey: topology.kubernetes.io/zone
10
      whenUnsatisfiable: DoNotSchedule
11
      labelSelector:
12
        matchLabels:
13
          app: nginx
14
    - maxSkew: 1
15
      topologyKey: kubernetes.io/hostname
16
      whenUnsatisfiable: ScheduleAnyway
17
      labelSelector:
18
        matchLabels:
19
          app: nginx

1
# 添加污点
2
kubectl taint nodes node1 key=value:NoSchedule
3
kubectl taint nodes node1 dedicated=gpu:NoExecute
4
kubectl taint nodes node1 temporary=true:NoSchedule
5

6
# 查看污点
7
kubectl describe node node1 | grep Taints
8

9
# 删除污点
10
kubectl taint nodes node1 key=value:NoSchedule-
11
kubectl taint nodes node1 dedicated-

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: gpu-app
5
spec:
6
  tolerations:
7
    # 匹配任意值的污点
8
    - key: "dedicated"
9
      operator: "Exists"
10
      effect: "NoSchedule"
11

12
    # 匹配特定值
13
    - key: "dedicated"
14
      operator: "Equal"
15
      value: "gpu"
16
      effect: "NoSchedule"
17

18
    # 匹配所有污点
19
    - key: "node.kubernetes.io/not-ready"
20
      operator: "Exists"
21
      effect: "NoExecute"
22
      tolerationSeconds: 300
23

24
    # 无视所有污点
25
    - operator: "Exists"

1
apiVersion: v1
2
kind: ResourceQuota
3
metadata:
4
  name: compute-quota
5
  namespace: default
6
spec:
7
  hard:
8
    requests.cpu: "10"
9
    requests.memory: 20Gi
10
    limits.cpu: "20"
11
    limits.memory: 40Gi
12
    pods: "50"
13
    services: "10"
14
    persistentvolumeclaims: "5"

1
# 查看配额使用
2
kubectl get resourcequota -o wide
3
kubectl describe resourcequota compute-quota

1
apiVersion: v1
2
kind: LimitRange
3
metadata:
4
  name: limits
5
  namespace: default
6
spec:
7
  limits:
8
    - type: Container
9
      max:
10
        cpu: "2"
11
        memory: 1Gi
12
      min:
13
        cpu: "100m"
14
        memory: 64Mi
15
      default:
16
        cpu: "500m"
17
        memory: 256Mi
18
      defaultRequest:
19
        cpu: "200m"
20
        memory: 128Mi
21
      maxLimitRequestRatio:
22
        cpu: "10"
23
        memory: "4"

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: nginx
5
spec:
6
  containers:
7
    - name: nginx
8
      image: nginx:1.25
9
      resources:
10
        requests:
11
          memory: "64Mi"
12
          cpu: "250m"
13
        limits:
14
          memory: "128Mi"
15
          cpu: "500m"

1
# 资源请求 (requests) - 调度依据
2
# 资源限制 (limits) - 限制最大使用
3

4
# QoS 等级
5
# 1. Guaranteed ( limits == requests )
6
# 2. Burstable ( requests < limits )
7
# 3. BestEffort ( 无 requests 和 limits )

1
apiVersion: scheduling.k8s.io/v1
2
kind: PriorityClass
3
metadata:
4
  name: high-priority
5
value: 100000
6
globalDefault: false
7
description: "高优先级 Pod"

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: important-app
5
spec:
6
  priorityClassName: high-priority
7
  containers:
8
    - name: app
9
      image: app:latest

1
apiVersion: kubescheduler.config.k8s.io/v1beta3
2
kind: KubeSchedulerConfiguration
3
profiles:
4
  - pluginConfig:
5
      - name: NodeResources
6
        args:
7
          mode: Least

1
# 使用自定义调度器
2
kubectl create configmap my-scheduler-config \
3
  --from-file=kube-scheduler-config.yaml \
4
  -n kube-system
5

6
# 创建使用自定义调度器的 Pod
7
kubectl create -f pod.yaml

1
# 为系统组件预留资源
2
# kube-apiserver: 至少 500m CPU, 256Mi 内存
3
# kubelet: 至少 500m CPU, 500Mi 内存

1
# 调度器考虑的因
2
# 1. 资源需求（CPU、内存、GPU）
3
# 2. 亲和性/反亲和性规则
4
# 3. 污点与容忍
5
# 4. 优先级
6
# 5. 拓扑分布约束
7
# 6. 污点容忽时间