如何配置topologySpreadConstraints固定K8s Pod在指定可用区的分布？

要实现Pod始终固定在zone-a运行3个、zone-b运行2个，仅靠topologySpreadConstraints做不到——它的作用是维持副本分布的最大偏差（这里maxSkew=1允许差1），但没法强制指定每个zone的精确副本数。最可靠的方式是把原Deployment拆分为两个独立的Deployment，分别绑定到目标zone并设置对应副本数：

方法：拆分Deployment + Node Affinity #

1. 针对zone-a的Deployment（3副本） #

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment-zone-a
  labels:
    app: nginx
    zone: a
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
      zone: a
  template:
    metadata:
      labels:
        app: nginx
        zone: a
    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
            - matchExpressions:
              - key: topology.kubernetes.io/zone
                operator: In
                values:
                - a
      containers:
      - name: nginx
        image: nginx:1.14.2
        ports:
        - containerPort: 80

2. 针对zone-b的Deployment（2副本） #

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment-zone-b
  labels:
    app: nginx
    zone: b
spec:
  replicas: 2
  selector:
    matchLabels:
      app: nginx
      zone: b
  template:
    metadata:
      labels:
        app: nginx
        zone: b
    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
            - matchExpressions:
              - key: topology.kubernetes.io/zone
                operator: In
                values:
                - b
      containers:
      - name: nginx
        image: nginx:1.14.2
        ports:
        - containerPort: 80

为什么不继续用topologySpreadConstraints？ #

topologySpreadConstraints的设计目标是均衡分布，而非精确配额。当总副本数无法被zone数量整除时（比如5个副本分2个zone），它只会保证各zone的副本数差距不超过maxSkew，但不会强制指定哪个zone多一个。拆分Deployment的方式能直接锁定每个zone的副本数，完全符合你的需求。

内容的提问来源于stack exchange，提问作者Boris