I deployed a K8s StatefulSet with 30 replicas (or N replicas, where N is multiple of 3) in EKS Cluster.
EKS cluster is with 3 nodes, one node for one AZ, and I want to guarantee with Kubernetes Affinity/AntiAffinity the equal distribution of pods across different AZ.
us-west-2a (n nodes) -> N/3 pods
us-west-2b (m nodes) -> N/3 pods
us-west-2c (o nodes) -> N/3 podsThanks
You can always use selectors and default labels which you get from AWS. A simple pod spec example is here
topologySpreadConstraints:
- maxSkew: 1
topologyKey: "topology.kubernetes.io/zone"
whenUnsatisfiable: DoNotSchedule
labelSelector:
matchLabels:
app: myappYou can include skew and other options based on the need as described here: https://kubernetes.io/docs/concepts/workloads/pods/pod-topology-spread-constraints/
While this is too possible with node affinity, a straight forward way is the use of topologySpreadContraints, here's the k8s documentation, diagrams and examples to see it in action.