基于集群中节点数量的水平 Pod Autoscaler 副本

Question

我正在寻找能够在节点加入集群时自动扩展 Pod 并在删除节点时缩回的解决方案。 我们在节点上运行 WebApp，当节点被安排断开连接时，这需要优雅的 pod 驱逐/终止。 我正在检查使用 DaemonSet 的选项，但由于我们使用 Kops 进行集群滚动更新，它忽略了 DaemonSets 驱逐（不支持标志“--ignore-daemionset”）。 结果，WebApp 与我们的应用程序不可接受的节点一起“死亡”。 Horizo​​ntalPodAutoscaler 覆盖部署 yaml 中设置的副本数量的能力可以解决这个问题。 我想找到根据集群中节点数量动态更改 Horizo​​ntalPodAutoscaler yaml 中的 min/maxReplicas 的方法。

spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: MyWebApp
  minReplicas: "Num of nodes in the cluster"
  maxReplicas: "Num of nodes in the cluster"

任何想法如何获取节点数量并相应地更新集群中的 Horizo​​ntalPodAutoscaler yaml？或者任何其他解决问题的方法？

Answer 1

您是否尝试过在 daemonset yaml 中使用 nodeSelector 规范。 因此，如果您在 yaml 中设置了节点选择器，并且如果您从节点中删除节点选择器标签值，则在耗尽之前，守护程序集应该优雅地缩小当您将新节点添加到集群标签具有自定义值的节点时也是如此并且 deamonset 将扩大规模。

这对我有用，所以你可以试试这个并与 Kops 确认

首先：使用集群上始终拥有的自定义标签标记所有节点

例子：

kubectl label nodes k8s-master-1 mylabel=allow_demon_set  
kubectl label nodes k8s-node-1 mylabel=allow_demon_set
kubectl label nodes k8s-node-2 mylabel=allow_demon_set
kubectl label nodes k8s-node-3 mylabel=allow_demon_set

然后到你的守护进程集 yaml 添加节点选择器。

Example.yaml 使用如下：注意 添加节点选择器字段

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: fluentd-elasticsearch
  labels:
    k8s-app: fluentd-logging
spec:
  selector:
    matchLabels:
      name: fluentd-elasticsearch
  template:
    metadata:
      labels:
        name: fluentd-elasticsearch
    spec:
      nodeSelector:
        mylabel: allow_demon_set
      tolerations:
      - key: node-role.kubernetes.io/master
        effect: NoSchedule
      containers:
      - name: fluentd-elasticsearch
        image: quay.io/fluentd_elasticsearch/fluentd:v2.5.2
        resources:
          limits:
            memory: 200Mi
          requests:
            cpu: 100m
            memory: 200Mi
        volumeMounts:
        - name: varlog
          mountPath: /var/log
        - name: varlibdockercontainers
          mountPath: /var/lib/docker/containers
          readOnly: true
      terminationGracePeriodSeconds: 30
      volumes:
      - name: varlog
        hostPath:
          path: /var/log
      - name: varlibdockercontainers
        hostPath:
          path: /var/lib/docker/containers

所以节点标记如下

$ kubectl get nodes --show-labels
NAME           STATUS   ROLES    AGE   VERSION   LABELS
k8s-master-1   Ready    master   9d    v1.17.0   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master-1,kubernetes.io/os=linux,mylable=allow_demon_set,node-role.kubernetes.io/master=
k8s-node-1     Ready    <none>   9d    v1.17.0   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node-1,kubernetes.io/os=linux,mylable=allow_demon_set
k8s-node-2     Ready    <none>   9d    v1.17.0   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node-2,kubernetes.io/os=linux,mylable=allow_demon_set
k8s-node-3     Ready    <none>   9d    v1.17.0   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node-3,kubernetes.io/os=linux,mylable=allow_demon_set

一旦你有正确的 yaml 使用它启动守护程序集

$ kubectl create -f Example.yaml

$ kubectl get all -o wide
NAME                              READY   STATUS    RESTARTS   AGE   IP            NODE           NOMINATED NODE   READINESS GATES
pod/fluentd-elasticsearch-jrgl6   1/1     Running   0          20s   10.244.3.19   k8s-node-3     <none>           <none>
pod/fluentd-elasticsearch-rgcm2   1/1     Running   0          20s   10.244.0.6    k8s-master-1   <none>           <none>
pod/fluentd-elasticsearch-wccr9   1/1     Running   0          20s   10.244.1.14   k8s-node-1     <none>           <none>
pod/fluentd-elasticsearch-wxq5v   1/1     Running   0          20s   10.244.2.33   k8s-node-2     <none>           <none>

NAME                 TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE   SELECTOR
service/kubernetes   ClusterIP   10.96.0.1    <none>        443/TCP   9d    <none>

NAME                                   DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR             AGE   CONTAINERS              IMAGES                                         SELECTOR
daemonset.apps/fluentd-elasticsearch   4         4         4       4            4           mylable=allow_demon_set   20s   fluentd-elasticsearch   quay.io/fluentd_elasticsearch/fluentd:v2.5.2   name=fluentd-elasticsearch

然后在排空节点之前，我们可以从节点中删除自定义标签，并且 pod-应该优雅地缩小，然后排空节点。

$ kubectl label nodes k8s-node-3 mylabel-

检查守护程序集，它应该缩小

ubuntu@k8s-kube-client:~$ kubectl get all -o wide
NAME                              READY   STATUS        RESTARTS   AGE     IP            NODE           NOMINATED NODE   READINESS GATES
pod/fluentd-elasticsearch-jrgl6   0/1     Terminating   0          2m36s   10.244.3.19   k8s-node-3     <none>           <none>
pod/fluentd-elasticsearch-rgcm2   1/1     Running       0          2m36s   10.244.0.6    k8s-master-1   <none>           <none>
pod/fluentd-elasticsearch-wccr9   1/1     Running       0          2m36s   10.244.1.14   k8s-node-1     <none>           <none>
pod/fluentd-elasticsearch-wxq5v   1/1     Running       0          2m36s   10.244.2.33   k8s-node-2     <none>           <none>

NAME                 TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE   SELECTOR
service/kubernetes   ClusterIP   10.96.0.1    <none>        443/TCP   9d    <none>

NAME                                   DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR             AGE     CONTAINERS              IMAGES                                         SELECTOR
daemonset.apps/fluentd-elasticsearch   3         3         3       3            3           mylable=allow_demon_set   2m36s   fluentd-elasticsearch   quay.io/fluentd_elasticsearch/fluentd:v2.5.2   name=fluentd-elasticsearch

现在再次将标签添加到具有相同自定义标签的新节点，当它添加到集群时，deamonset 将扩大规模

$ kubectl label nodes k8s-node-3 mylable=allow_demon_set

ubuntu@k8s-kube-client:~$ kubectl get all -o wide
NAME                              READY   STATUS    RESTARTS   AGE     IP            NODE           NOMINATED NODE   READINESS GATES
pod/fluentd-elasticsearch-22rsj   1/1     Running   0          2s      10.244.3.20   k8s-node-3     <none>           <none>
pod/fluentd-elasticsearch-rgcm2   1/1     Running   0          5m28s   10.244.0.6    k8s-master-1   <none>           <none>
pod/fluentd-elasticsearch-wccr9   1/1     Running   0          5m28s   10.244.1.14   k8s-node-1     <none>           <none>
pod/fluentd-elasticsearch-wxq5v   1/1     Running   0          5m28s   10.244.2.33   k8s-node-2     <none>           <none>

NAME                 TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE   SELECTOR
service/kubernetes   ClusterIP   10.96.0.1    <none>        443/TCP   9d    <none>

NAME                                   DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR             AGE     CONTAINERS              IMAGES                                         SELECTOR
daemonset.apps/fluentd-elasticsearch   4         4         4       4            4           mylable=allow_demon_set   5m28s   fluentd-elasticsearch   quay.io/fluentd_elasticsearch/fluentd:v2.5.2   name=fluentd-elasticsearch

请确认您是否想要这样做并与 kops 合作