Kubernetes Pod 调度全解析：从 nodeSelector 到亲和性调度

1. 过滤（Predicate）：筛掉不满足条件的节点
2. 打分（Priority）：对候选节点评分
3. 选定：得分最高的节点，binding

apiVersion: v1
kind: Pod
metadata:
  name: nginx
spec:
  nodeName: node-1      # 直接指定，跳过调度器
  containers:
  - name: nginx
    image: nginx:1.25

# 给节点打标签
kubectl label nodes node-1 disktype=ssd
kubectl label nodes node-2 zone=beijing

spec:
  nodeSelector:
    disktype: ssd

spec:
  affinity:
    nodeAffinity:
      # 硬约束：必须满足
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key: kubernetes.io/os
            operator: In
            values: [linux]
      # 软约束：尽量满足，权重 1-100
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 80
        preference:
          matchExpressions:
          - key: disktype
            operator: In
            values: [ssd]

spec:
  affinity:
    # 反亲和：同一应用不要调度到同一节点（高可用）
    podAntiAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - key: app
            operator: In
            values: [web]
        topologyKey: kubernetes.io/hostname

    # 亲和：与 redis 部署在同一可用区
    podAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        podAffinityTerm:
          labelSelector:
            matchExpressions:
            - key: app
              operator: In
              values: [redis]
          topologyKey: topology.kubernetes.io/zone

# 给节点加污点（禁止普通 Pod 调度）
kubectl taint nodes node-1 dedicated=gpu:NoSchedule

# 移除
kubectl taint nodes node-1 dedicated=gpu:NoSchedule-

spec:
  tolerations:
  - key: "dedicated"
    operator: "Equal"
    value: "gpu"
    effect: "NoSchedule"
    tolerationSeconds: 3600

spec:
  topologySpreadConstraints:
  - maxSkew: 1
    topologyKey: kubernetes.io/hostname
    whenUnsatisfiable: DoNotSchedule
    labelSelector:
      matchLabels:
        app: web

resources:
  requests:
    cpu: "500m"
    memory: "512Mi"
  limits:
    cpu: "2"
    memory: "2Gi"

kubectl describe pod <pod> | grep -A 10 Events
# 常见原因：
# - 资源不足
# - 节点污点未容忍
# - PVC 未绑定
# - 亲和性规则无法满足