不同namespace的pod反亲和配置

喜欢ヅ旅行 2022-09-04 07:59 116阅读 0赞

### 业务需求 两pod互斥，增加系统健壮性 ###

pod1:  
\-name:prometheus-dep  
\-namespace:prometheus

pod2:  
\-name:prometheus-server  
\-namespace:prometheus-server-dep

测试结论  
查看两个pod的label 备用

kubectl -n <ns> describe pods  <pods-name> |grep Labels

pod1.yaml  
配置模块均配置对方pod的信息

apiVersion: apps/v1
    kind: Deployment
    metadata:
      name: mynginx-1 
      namespace: dev-1
    spec:
      replicas: 2
      selector:
        matchLabels:
          app: mynginx-1 
      template:
        metadata:
          labels:
            app: mynginx-1 
          containers:
          - name: mycontainer-1
            image: nginx
            imagePullPolicy: IfNotPresent
            ports:
            - name: nginx
              containerPort: 80
            volumeMounts:
            - mountPath: "/usr/share/nginx/html/"
              name: nginx-vol
          affinity: ##调度策略配置模块
            podAntiAffinity:   ##类型pod反亲和
              requiredDuringSchedulingIgnoredDuringExecution: ##硬需求
              - labelSelector: ##标签匹配
                  matchExpressions:
                  - key: app  ##刚刚查到另一个pod的标签 key值
                    operator: In  ##匹配规则
                    values: [mynginx-test] ##对应的value值
                topologyKey: kubernetes.io/hostname ##固定字段控制域
                namespaces: [dev-2] ##互斥pod的名存空间

pod2.yaml

apiVersion: apps/v1
    kind: Deployment
    metadata:
      name: mynginx-test 
      namespace: dev-2
    spec:
      replicas: 2
      selector:
        matchLabels:
          app: mynginx-test 
      template:
        metadata:
          labels:
            app: mynginx-test 
        spec:
          containers:
          - name: mycontainer-2
            image: nginx
            imagePullPolicy: IfNotPresent
            ports:
            - name: nginx
              containerPort: 80
            volumeMounts:
            - mountPath: "/usr/share/nginx/html/"
              name: nginx-vol
          affinity:
            podAntiAffinity:
              requiredDuringSchedulingIgnoredDuringExecution:
              - labelSelector: 
                  matchExpressions:
                  - key: app ##对应标签的key
                    operator: In
                    values: [mynginx-1] ##对应标签的value
                topologyKey: kubernetes.io/hostname 
                namespaces: [dev-1] ##对方pod的名称空间
          volumes:
          - name: nginx-vol
            hostPath:
              path: /data/svc/web1-html

这样两个pod都配置互斥的好处是:不会因为启动顺序而失效，无论谁先启动都会检测 分配到合适的node

**注意：唯一缺陷**

当dep启动的pod 副本数大于或等于 集群 node数量时，可能会出现每个pod分布在每个node节点上的情况，导致另一个pod 不能分配工作节点，一直处于pending状态