k8s----節點親和性和污點容忍

1、查看標簽
kubectl  get nodes --show-labels

2、打標簽
kubectl label nodes node1.com kong=true 
kubectl label nodes node2.com kong=true 
kubectl label nodes node3.com kong=true

3、node添加親和性
spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
            - matchExpressions:
              - key: kong
                operator: In
                values:
                - "true"
4、刪除label
kubectl label nodes node1.com kong-

 

反親和性

affinity:
  podAntiAffinity:
    preferredDuringSchedulingIgnoredDuringExecution:
    - weight: 100
      podAffinityTerm:
        labelSelector:
          matchExpressions:
           - key: k8s-app
             operator: In
             values:
             - my-app
        topologyKey: kubernetes.io/hostname

 

 

節點綁定 
    spec:
      containers:
      dnsPolicy: ClusterFirst
      nodeSelector:
        kubernetes.io/hostname: node03.paas

 

污點和容忍

NoSchedule：表示 pod 不會被調度到標記為 taints 的節點，只會影響新的 pod 調度
PreferNoSchedule：NoSchedule 的軟策略版本，表示盡量不調度到污點節點上去，只會影響新的 pod 調度
NoExecute：該選項意味着一旦 Taint 生效，如該節點內正在運行的 pod 沒有對應 Tolerate 設置，會直接被逐出

# 節點添加污點
$ kubectl taint nodes node02 app=heihei:NoSchedule
# 取消節點污點
 kubectl taint nodes node02 app-
# 查看節點污點
kubectl describe nodes node02|grep Taints

 
如果 operator 的值是 Exists，則 value 屬性可省略
如果 operator 的值是 Equal，則表示其 key 與 value 之間的關系是 equal(等於)
如果不指定 operator 屬性，則默認值為 Equal

如果pod想部署到此節點，需要添加如下內容
spec:
      containers:
      - name: nginx
        image: nginx:1.7.9
      tolerations:
      - key: "app"
        operator: "Exists"
        effect: "NoSchedule" #這里的值要和上面的的添加污點的值匹配上，不然調度不到

重要：如果只調度到此污點節點，pod還需要添加如下親和性
kubectl  label node yq-map app=my-map

affinity:
  podAntiAffinity:
    preferredDuringSchedulingIgnoredDuringExecution:
    - weight: 100 podAffinityTerm: labelSelector: matchExpressions: - key: app operator: In values: - my-app

 參考：https://blog.csdn.net/qq_34556414/article/details/109448632