k8s的亲和性设置，指定节点，分散部署pod，捆绑部署pod

前言：

亲和性是affinity，分为两种

-- 强制：requiredDuringSchedulingIgnoredDuringExecution

-- 参考：preferredDuringSchedulingIgnoredDuringExecution

// 也就第一个词不同了，IgnoredDuringExecution的意思是：部署完了就不再管这些规则。比如部署后我把node的标签改了导致不满足规则了，会重新调度吗？不会，根本不去检测变化。

也分三种：

-- 亲和节点：nodeAffinity

-- 亲和pod：podAffinity

-- 不亲和pod：podAntiAffinity

另外：

部署pod一定是到node上的

-- 亲和某个node，到某个node上

-- 亲和pod，是指亲和（某些节点上的）pod，到 某些节点 上。到底是哪些节点？用topologyKey指定。

一、指定节点

1.强制指定

（1）简单点的设置可以用nodeName和nodeSelector

（2）使用亲和性nodeAffinity

2.尽量满足 -- 意思是可以不满足

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test
spec:
  selector:
    matchLabels:
      app: test
  replicas: 1
  template:
    metadata:
      labels:
        app: test
    spec:
      # nodeName: master      #（1/3）指定node名称
      # nodeSelector:         #（2/3）指定node标签
        # kubernetes.io/hostname: master
      affinity:             #（3/3）使用亲和性
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:  # 强制满足
            nodeSelectorTerms:     # terms里满足一个就够了（或的关系）
              - matchExpressions:  # match里要全部满足 （且的关系）
                  - key: kubernetes.io/hostname
                    operator: In
                    values:
                     - master
                  # - key: kubernetes.io/hostname
                    # operator: NotIn
                    # values:
                     # - master
              - matchExpressions:
                  - key: kubernetes.io/hostname
                    operator: NotIn
                    values:
                     - master
          preferredDuringSchedulingIgnoredDuringExecution:  # 尽量满足，可以不满足
            - weight: 10
              preference:
                matchExpressions:
                  - key: kubernetes.io/hostname
                    operator: In
                    values:
                      - abcd
      restartPolicy: Always
      containers:
        - name: busybox
          image: busybox
          imagePullPolicy: IfNotPresent
          command:
            - sleep
            - "3600"

二、分散部署pod

比如部署redis、minio集群，希望pod尽量不在同一个节点

使用亲和性podAntiAffinity

1.强制要求

2.尽量满足

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test
spec:
  selector:
    matchLabels:
      app: test
  replicas: 3
  template:
    metadata:
      labels:
        app: test
    spec:
      # nodeName: master      #（1/3）指定node名称
      # nodeSelector:         #（2/3）指定node标签
        # kubernetes.io/hostname: master
      affinity:             #（3/3）使用亲和性
        podAntiAffinity:
          # requiredDuringSchedulingIgnoredDuringExecution:
            # - topologyKey: kubernetes.io/hostname  # topologyKey用来筛选node，确定一个比较的范围
              # labelSelector:
                # matchExpressions:
                  # - key: app
                    # operator: In
                    # values:
                      # - test
          preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 10
              podAffinityTerm:
                topologyKey: kubernetes.io/hostname
                labelSelector:
                  matchExpressions:
                    - key: app
                      operator: In
                      values:
                        - test
        # nodeAffinity:
          # requiredDuringSchedulingIgnoredDuringExecution:  # 强制满足
            # nodeSelectorTerms:     # terms里满足一个就够了（或的关系）
              # - matchExpressions:  # match里要全部满足 （且的关系）
                  # - key: kubernetes.io/hostname
                    # operator: In
                    # values:
                     # - master
              # - matchExpressions:
                  # - key: kubernetes.io/hostname
                    # operator: NotIn
                    # values:
                     # - master
          # preferredDuringSchedulingIgnoredDuringExecution:  # 尽量满足，可以不满足
            # - weight: 10
              # preference:
                # matchExpressions:
                  # - key: kubernetes.io/hostname
                    # operator: In
                    # values:
                      # - abcd
      restartPolicy: Always
      containers:
        - name: busybox
          image: busybox
          imagePullPolicy: IfNotPresent
          command:
            - sleep
            - "3600"

三、捆绑部署pod

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test
spec:
  selector:
    matchLabels:
      app: test
  replicas: 3
  template:
    metadata:
      labels:
        app: test
    spec:
      # nodeName: master      #（1/3）指定node名称
      # nodeSelector:         #（2/3）指定node标签
        # kubernetes.io/hostname: master
      affinity:             #（3/3）使用亲和性
        podAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            - topologyKey: kubernetes.io/hostname  # topologyKey用来筛选node，确定一个比较的范围
              labelSelector:
                matchExpressions:
                  - key: app
                    operator: In
                    values:
                      - test2
          preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 10
              podAffinityTerm:
                topologyKey: kubernetes.io/hostname
                labelSelector:
                  matchExpressions:
                    - key: app
                      operator: In
                      values:
                        - test2
        # podAntiAffinity:
          # requiredDuringSchedulingIgnoredDuringExecution:
            # - topologyKey: kubernetes.io/hostname  
              # labelSelector:
                # matchExpressions:
                  # - key: app
                    # operator: In
                    # values:
                      # - test
          # preferredDuringSchedulingIgnoredDuringExecution:
            # - weight: 10
              # podAffinityTerm:
                # topologyKey: kubernetes.io/hostname
                # labelSelector:
                  # matchExpressions:
                    # - key: app
                      # operator: In
                      # values:
                        # - test
        # nodeAffinity:
          # requiredDuringSchedulingIgnoredDuringExecution:  # 强制满足
            # nodeSelectorTerms:     # terms里满足一个就够了（或的关系）
              # - matchExpressions:  # match里要全部满足 （且的关系）
                  # - key: kubernetes.io/hostname
                    # operator: In
                    # values:
                     # - master
              # - matchExpressions:
                  # - key: kubernetes.io/hostname
                    # operator: NotIn
                    # values:
                     # - master
          # preferredDuringSchedulingIgnoredDuringExecution:  # 尽量满足，可以不满足
            # - weight: 10
              # preference:
                # matchExpressions:
                  # - key: kubernetes.io/hostname
                    # operator: In
                    # values:
                      # - abcd
      restartPolicy: Always
      containers:
        - name: busybox
          image: busybox
          imagePullPolicy: IfNotPresent
          command:
            - sleep
            - "3600"

目录：https://blog.csdn.net/u013595395/article/details/114527658