Container Service for Kubernetes:Best practices for associating deployment sets with node pools

Two layers of high availability

Deployment sets and availability zones address different failure domains. You can use them independently or together:

Example 2 in the Examples section below combines both layers.

Example 1: One pod per node using pod anti-affinity

This example schedules three replicas of an nginx Deployment to three different nodes within the same node pool.

1. Create a node pool with three nodes and associate it with a deployment set. See Associate a deployment set with a node pool.

2. On the Node Pools page, click the node pool name. On the Nodes tab, confirm that three nodes appear.

3. Log on to the ECS console. In the left-side navigation pane, choose Deployment & Elasticity > Deployment Sets. Confirm that all three nodes belong to the specified deployment set.

4. Apply the following YAML. The podAntiAffinity rule with topologyKey: kubernetes.io/hostname ensures no two pods from the same Deployment land on the same node. The nodeSelector pins pods to the target node pool.

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: nginx
     labels:
       app: nginx
   spec:
     replicas: 3
     selector:
       matchLabels:
         app: nginx
     template:
       metadata:
         name: nginx
         labels:
           app: nginx
       spec:
         affinity:
           podAntiAffinity:
             requiredDuringSchedulingIgnoredDuringExecution:  # Hard anti-affinity: pods must land on distinct nodes.
             - labelSelector:
                 matchExpressions:
                 - key: app
                   operator: In
                   values:
                   - nginx
               topologyKey: kubernetes.io/hostname
         nodeSelector:
             alibabacloud.com/nodepool-id: <nodepool-id>  # Replace with your node pool ID.
         containers:
         - name: nginx
           image: anolis-registry.cn-zhangjiakou.cr.aliyuncs.com/openanolis/nginx:1.14.1-8.6
           resources:
             limits:
               cpu: 1
             requests:
               cpu: 1

Result: On the Deployments page, click the Deployment. On the Pods tab, confirm that each pod is scheduled to a different node.

Example 2: Pods spread across multiple zones using topology spread constraints

This example schedules four replicas of an nginx Deployment across four nodes in four different availability zones. It combines deployment set isolation (physical server level) with multi-zone vSwitches (availability zone level) for two layers of high availability.

1. Create a node pool with four nodes, associate it with a deployment set, and select vSwitches in multiple zones. See Associate a deployment set with a node pool.

2. On the Node Pools page, click the node pool name. On the Nodes tab, confirm that four nodes across four zones appear. Auto Scaling uses a balanced distribution policy to place ECS instances evenly across zones.

3. Log on to the ECS console. In the left-side navigation pane, choose Deployment & Elasticity > Deployment Sets. Confirm that all four nodes belong to the specified deployment set.

4. Apply the following YAML. The two topologySpreadConstraints entries enforce even spread by both hostname and zone, with DoNotSchedule preventing pods from concentrating on fewer nodes or zones. For more information, see Pod topology spread constraints.

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: nginx
     labels:
       app: nginx
   spec:
     replicas: 4
     selector:
       matchLabels:
         app: nginx
     template:
       metadata:
         name: nginx
         labels:
           app: nginx
       spec:
         topologySpreadConstraints:
           - maxSkew: 1
             topologyKey: kubernetes.io/hostname       # Spread evenly across nodes.
             whenUnsatisfiable: DoNotSchedule
             labelSelector:
                 matchLabels:
                   app: nginx
           - maxSkew: 1
             topologyKey: topology.kubernetes.io/zone  # Spread evenly across zones.
             whenUnsatisfiable: DoNotSchedule
             labelSelector:
                 matchLabels:
                   app: nginx
         nodeSelector:
             alibabacloud.com/nodepool-id: <nodepool-id>  # Replace with your node pool ID.
         containers:
         - name: nginx
           image: anolis-registry.cn-zhangjiakou.cr.aliyuncs.com/openanolis/nginx:1.14.1-8.6
           resources:
             limits:
               cpu: 1
             requests:
               cpu: 1

Result: On the Deployments page, click the Deployment. On the Pods tab, confirm that pods are distributed across different nodes in different zones.