您当前的位置： 首页 > 慢生活 > 程序人生 网站首页程序人生

49、集群调度 节点亲和性（1）

发布时间：2023-01-05 21:46:33编辑：雪饮阅读（）

Step1

首先来温习下标签

[root@k8s-master01 ~]# kubectl get node --show-labels

NAME           STATUS   ROLES    AGE   VERSION   LABELS

k8s-master01   Ready    master   22d   v1.15.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master01,kubernetes.io/os=linux,node-role.kubernetes.io/master=

k8s-node01     Ready    <none>   22d   v1.15.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01,kubernetes.io/os=linux

k8s-node02     Ready    <none>   22d   v1.15.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02,kubernetes.io/os=linux

可以看到这里有两个特殊的标签

kubernetes.io/hostname=k8s-node01

kubernetes.io/hostname=k8s-node02

然后接下来咱们先清理下之前的pod

[root@k8s-master01 affi]# kubectl delete pod --all

pod "test-pd" deleted

然后创建自己的pod

[root@k8s-master01 affi]# cat pod1.yaml

apiVersion: v1

kind: Pod

metadata:

 name: affinity

 labels:

  app: node-affinity-pod

spec:

 containers:

 - name: with-node-affinity

   image: wangyanglinux/myapp:v1

 affinity:

  nodeAffinity:

   requiredDuringSchedulingIgnoredDuringExecution:

    nodeSelectorTerms:

    - matchExpressions:

      - key: kubernetes.io/hostname

        operator: NotIn

        values:

        - k8s-node02

硬性亲和性，要求不在节点2上。

[root@k8s-master01 affi]# kubectl create -f pod1.yaml

pod/affinity created

那么果然他在节点1上

[root@k8s-master01 affi]# kubectl get pod -o wide

NAME       READY   STATUS    RESTARTS   AGE    IP             NODE         NOMINATED NODE   READINESS GATES

affinity   1/1     Running   0          2m7s   10.224.1.188   k8s-node01   <none>           <none>

经过多次测试，仍旧绑定在节点1上

[root@k8s-master01 affi]# kubectl delete pod --all && kubectl create -f pod1.yaml && kubectl get pod -o wide

pod "affinity" deleted

pod/affinity created

NAME       READY   STATUS              RESTARTS   AGE   IP       NODE         NOMINATED NODE   READINESS GATES

affinity   0/1     ContainerCreating   0          0s    <none>   k8s-node01   <none>           <none>

[root@k8s-master01 affi]# kubectl delete pod --all && kubectl create -f pod1.yaml && kubectl get pod -o wide

pod "affinity" deleted

pod/affinity created

NAME       READY   STATUS              RESTARTS   AGE   IP       NODE         NOMINATED NODE   READINESS GATES

affinity   0/1     ContainerCreating   0          0s    <none>   k8s-node01   <none>           <none>

[root@k8s-master01 affi]# kubectl delete pod --all && kubectl create -f pod1.yaml && kubectl get pod -o wide

pod "affinity" deleted

pod/affinity created

NAME       READY   STATUS              RESTARTS   AGE   IP       NODE         NOMINATED NODE   READINESS GATES

affinity   0/1     ContainerCreating   0          0s    <none>   k8s-node01   <none>           <none>

都不会在node02上

这就是节点的硬亲和性策略

Step2

再次将pod清理下

[root@k8s-master01 affi]# kubectl delete pod --all

pod "affinity" deleted

然后这次咱们用In，也就是只让其在node02节点上

[root@k8s-master01 affi]# cat pod2.yaml

apiVersion: v1

kind: Pod

metadata:

 name: affinity

 labels:

  app: node-affinity-pod

spec:

 containers:

 - name: with-node-affinity

   image: wangyanglinux/myapp:v1

 affinity:

  nodeAffinity:

   requiredDuringSchedulingIgnoredDuringExecution:

    nodeSelectorTerms:

    - matchExpressions:

      - key: kubernetes.io/hostname

        operator: In

        values:

        - k8s-node02

可以看到就很不错，的确只在node02上

[root@k8s-master01 affi]# kubectl create -f pod2.yaml && kubectl get pod -o wide

pod/affinity created

NAME       READY   STATUS              RESTARTS   AGE   IP       NODE         NOMINATED NODE   READINESS GATES

affinity   0/1     ContainerCreating   0          1s    <none>   k8s-node02   <none>           <none>

[root@k8s-master01 affi]# kubectl get pod -o wide

NAME       READY   STATUS    RESTARTS   AGE   IP             NODE         NOMINATED NODE   READINESS GATES

affinity   1/1     Running   0          65s   10.224.2.200   k8s-node02   <none>           <none>

Step3

咱们再来清理一波

[root@k8s-master01 affi]# kubectl delete pod --all

pod "affinity" deleted

这次我们让其亲和到一个不存在的节点上

[root@k8s-master01 affi]# cat pod3.yaml

apiVersion: v1

kind: Pod

metadata:

 name: affinity

 labels:

  app: node-affinity-pod

spec:

 containers:

 - name: with-node-affinity

   image: wangyanglinux/myapp:v1

 affinity:

  nodeAffinity:

   requiredDuringSchedulingIgnoredDuringExecution:

    nodeSelectorTerms:

    - matchExpressions:

      - key: kubernetes.io/hostname

        operator: In

        values:

        - k8s-node03

由于node03一直不在，所以一直处于pending状态

[root@k8s-master01 affi]# kubectl create -f pod3.yaml && kubectl get pod -o wide

pod/affinity created

NAME       READY   STATUS    RESTARTS   AGE   IP       NODE     NOMINATED NODE   READINESS GATES

affinity   0/1     Pending   0          0s    <none>   <none>   <none>           <none>

[root@k8s-master01 affi]# kubectl get pod -o wide

NAME       READY   STATUS    RESTARTS   AGE   IP       NODE     NOMINATED NODE   READINESS GATES

affinity   0/1     Pending   0          13s   <none>   <none>   <none>           <none>

由于是硬策略，所以这里一直就非要等到创建在node03节点上。。。也就不断的pending状态上。。。

接下来我们再次清理下

[root@k8s-master01 affi]# kubectl delete pod --all

pod "affinity" deleted