NodeSelector 不适用于多节点池？

Question

TL;DR: NodeSelector ignores nodes from another NodePool. How to distribute pods across more NodePools using a label nodeSelector or other technique?

我有两个这样的节点池：

...
# Spot node pool
resource "azurerm_kubernetes_cluster_node_pool" "aks_staging_np_compute_spot" {
  name                  = "computespot"
  (...)
  vm_size               = "Standard_F8s_v2"
  max_count             = 2
  min_count             = 2
  (...)
  priority = "Spot"
  eviction_policy = "Delete"
  (...)
  node_labels = {
    "pool_type" = "compute"
  }

# Regular node pool
resource "azurerm_kubernetes_cluster_node_pool" "aks_staging_np_compute_base" {
  name                  = "computebase"
  (...)
  vm_size               = "Standard_F8s_v2"
  max_count             = 2
  min_count             = 2
  node_labels = {
    "pool_type" = "compute"
  }

两个池都部署在 AKS 中，所有节点都处于 OK 状态。请注意两件事：

两者都有标签 pool_type: compute
两者的大小与 Standard_F8s_v2

（我的集群中还有其他20个不同标签的节点并不重要。）

然后我得到了这样的部署（为简洁起见省略了不相关的行）：

apiVersion: apps/v1
kind: Deployment
metadata:
  (...)
spec:
  replicas: 4
  selector:
    matchLabels:
      app: myapp
  template:
    (...)
    spec:
      nodeSelector:
        pool_type: compute
      (...)
      containers:
        (...)

tolerations 中还有一个条目用于接受 Azure 现货实例。它显然有效。

tolerations:
        - key: "kubernetes.azure.com/scalesetpriority"
          operator: "Equal"
          value: "spot"
          effect: "NoSchedule"

问题是 应用程序只部署在一个节点池上（"computespot" 在这种情况下）并且从不接触另一个（computebase).即使各个节点的标签和大小相同。

2 pods 在 computespot 个节点上运行，每个节点一个节点。
第二个 pods 没有安排经典错误 0/24 nodes are available: 14 Insufficient cpu, 17 Insufficient memory, 4 node(s) didn't match node selector. 这绝对是谎言 因为我可以看到 computebase 节点空荡荡地坐在那里。

如何解决？

Answer 1

找到了使用 pod affinity 的解决方案。

spec:
      # This didn't work:
      #
      # nodeSelector:
      #   pool_type: compute 
      # 
      # But this does:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                - key: pool_type
                  operator: In
                  values:
                  - compute

我不知道原因，因为我们仍在处理一个标签。如果有人知道，请分享。

NodeSelector 不适用于多节点池？

NodeSelector does not work for multiple node pools?

kubernetes

kubernetes-pod

azure-aks

spot-instances