Kubernetes 之 Pod 之间的亲和性与反亲和性

Pod 亲和性定义

Pod 亲和性是指将 Pod 和 Pod 根据某些约束性条件捆绑到具有相同特性的工作节点上，例如同一区域、同一机房乃至同一机柜。这样 Pod和 Pod 可以更好的通信，解决 IO 瓶颈。它同样分为硬亲和性requiredDuringSchedulingIgnoredDuringExecution和软亲和性preferredDuringSchedulingIgnoredDuringExecution，同时亲和性下面还有一个拓扑键topologyKey，这是用来将 Pod 约束到跟跟随 Pod 同一个域上的。

Pod 亲和性使用

我们先定义一个被跟随的 Pod

yaml 复制代码

apiVersion: v1
kind: Pod
metadata:
  name: pod-affinity-bind
  namespace: default
  labels:
    app: nginx
spec:
  containers:
  - name: k8s-test
    image: k8s-test:v1.0
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 80

定义 Pod 去跟随第一个 Pod

yaml 复制代码

apiVersion: v1
kind: Pod
metadata:
  name: pod-affinity-track
  namespace: default
  labels:
    app: track
spec:
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - key: app
            operator: In
            values:
            - nginx
        topologyKey: kubernetes.io/hostname
  containers:
  - name: k8s-test
    image: k8s-test:v1.0
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 80

运行结果

root@k8s-master1:~# kubectl get pods -owide --show-labels
NAME                 READY   STATUS    RESTARTS   AGE   IP              NODE          NOMINATED NODE   READINESS GATES   LABELS
pod-affinity-bind    1/1     Running   0          25m   10.244.194.83   k8s-worker1   <none>           <none>            app=nginx
pod-affinity-track   1/1     Running   0          24s   10.244.194.84   k8s-worker1   <none>           <none>            app=track

Pod 反亲和性定义

Pod 反亲和性是与亲和性完全相反的定义，一旦发现它的约束条件匹配，那么这个 Pod 比与匹配 Pod 在不同的topologyKey上的工作节点上。它主要可以用来保障不是热点区域和不需要高性能响应但很重要的业务不受干扰。

Pod 反亲和性使用

yaml 复制代码

apiVersion: v1
kind: Pod
metadata:
  name: pod-affinity-anti
  namespace: default
  labels:
    app: anti
spec:
  affinity:
    podAntiAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - key: app
            operator: In
            values:
            - nginx
        topologyKey: kubernetes.io/hostname
  containers:
  - name: k8s-test
    image: k8s-test:v1.0
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 80

shell 复制代码

root@k8s-master1:~# kubectl get pods -owide                          
NAME                 READY   STATUS    RESTARTS   AGE     IP              NODE          NOMINATED NODE   READINESS GATES
pod-affinity-anti    1/1     Running   0          8m45s   10.244.126.15   k8s-worker2   <none>           <none>
pod-affinity-bind    1/1     Running   0          76m     10.244.194.83   k8s-worker1   <none>           <none>
pod-affinity-track   1/1     Running   0          51m     10.244.194.84   k8s-worker1   <none>           <none>