k8s 亲和性，污点，容忍度

最新推荐文章于 2024-11-12 20:46:17 发布

q_97

最新推荐文章于 2024-11-12 20:46:17 发布

阅读量312

点赞数 3

分类专栏： k8s入门到精通文章标签： kubernetes 容器云原生

本文链接：https://blog.csdn.net/m0_73950916/article/details/142437659

版权

k8s入门到精通专栏收录该内容

12 篇文章 0 订阅

订阅专栏

1、节点选择器

没有亲和性的话,就会随机的调度到node节点上面去
可以指定调度到node节点上面去

1、nodename角度

根据nodename来调度

busybox要有一个持续的进程才能运行

apiVersion: v1
kind: Pod
metadata:
  name: p1
  namespace: default
  labels:
    app: n1
spec:
  nodeName: node1
  containers:
  - name: tomocat1
    image: docker.io/library/tomcat:8.5-jre8-alpine
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 8080
  - name: busybox1
    image: docker.io/library/busybox:1.28
    imagePullPolicy: IfNotPresent
    command:
    - "/bin/sh"   #分配一个终端
    - "-c"   #命令后面执行的参数
    - "sleep 3600000"

#k8s内部进行访问，浏览器访问不了
[root@master qinhexing]# kubectl get pod -o wide
NAME   READY   STATUS    RESTARTS   AGE   IP               NODE    NOMINATED NODE   READINESS GATES
p1     2/2     Running   0          88s   10.244.166.140   node1   <none>           <none>
[root@master qinhexing]# curl 10.244.166.140:8080




#Pod 的许多字段（如 nodeName、容器命令和参数、卷等）在 Pod 创建后都是不可变的。
需要强制删除pod --force --grace-period=0


内部访问

2、nodeselector角度

根据node上面标签来进行调度

#调度到node上面有disk=node1，re=1的标签，发现调度不了
[root@master qinhexing]# cat nodeselector.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: tomcat1
  namespace: dev
spec:
  nodeSelector:
    disk: node1
    re: rrr
  containers:
  - name: t1
    image: docker.io/library/tomcat:8.5-jre8-alpine
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 8080

#发现node上面没有指定的标签，处于pending状态
#3个node不能匹配pod标签
[root@master qinhexing]# kubectl describe pod -n dev
Node-Selectors:              disk=node1
                             re=rrr
Events:
  Type     Reason            Age   From               Message
  ----     ------            ----  ----               -------
  Warning  FailedScheduling  75s   default-scheduler  0/3 nodes are available: 1 node(s) had untolerated taint {node-role.kubernetes.io/control-plane: }, 3 node(s) didn't match Pod's node affinity/selector. preemption: 0/3 nodes are available: 3 Preemption is not helpful for scheduling..


#打上标签，发现Pod进行调度了
[root@master qinhexing]# kubectl label nodes node1 disk=node1 re=rrr
node/node1 labeled
[root@master qinhexing]# kubectl get pod -n dev
NAME      READY   STATUS    RESTARTS   AGE
tomcat1   1/1     Running   0          4m5s

#pod调度到node1上面了
  Normal   Scheduled         50s    default-scheduler  Successfully assigned dev/tomcat1 to node1

2、node亲和性

#硬亲和性，node必须符合才行
matchexpression：node标签
[root@master qinhexing]# cat nodeaff-re.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nodeaff-r
  namespace: dev
spec:
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:   #匹配的是node上面的标签
          - key: a      #a=q1的node节点上面去
            operator: "In" 
            values:
            - q1 
  containers:
  - name: nodeaff-r
    image: docker.io/library/nginx:1.9.1
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 80



#软亲和性，尽量满足，没有的话，找其他的节点
[root@master qinhexing]# cat nodeaff-per.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nodeaff-pre
  namespace: dev
spec:
  containers:
  - name: nodeaff-pre
    image: docker.io/library/nginx
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 80
  affinity:
    nodeAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 80  #权重占80
        preference:
          matchExpressions:
          - key: a
            operator: "In"
            values:
            - q1

3、pod亲和性

根据pod与pod之间的关系，有相同特征的话，就调度到同一个节点上面，访问的话，速度就更加的快一点
互为依赖数据库和web服务器，一起工作
根据已有的pod上面的标签，来进行调度

#硬亲和性
[root@master qinhexing]# cat podaff-re.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: pod-first
  namespace: dev
  labels:
    a: qq1
spec:
  containers:
  - name: pod-first
    image: docker.io/library/nginx
    imagePullPolicy: IfNotPresent
---
apiVersion: v1
kind: Pod
metadata:
  name: podaff-second
  namespace: dev
spec:
  containers:
  - name: podaff-re
    image: docker.io/ikubernetes/myapp:v1
    imagePullPolicy: IfNotPresent
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - topologyKey: kubernetes.io/hostname  #根据主机名来判断2个pod是否在同一个节点上面,也可以是自己的标签
        labelSelector:  #标签选择器
          matchExpressions:
          - key: a
            operator: "In"
            values:
            - qq1

#软亲和性

4、pod反亲和性

2个pod都占用很大的资源，使用反亲和性，就不会调用在同一个节点上面了


#to是一样的话，那么这个pod就不能进行调度了，因为判断的标签都是一样的则主机就是一样的，那么就不行
[root@master qinhexing]# cat podanti.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: pod1-anti
  namespace: dev
spec:
  containers:
  - name: pod1-anti
    image: docker.io/library/nginx
    imagePullPolicy: IfNotPresent
  affinity:
    podAntiAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - topologyKey: kubernetes.io/hostname 
        labelSelector:
          matchExpressions:
          - key: a
            operator: "In"
            values:
            - qq1

#需要注意的就是topologyKey，一样的话，那么就无法进行调度了，因为判断的节点都是一样的

5、污点和容忍度

- 污点就是节点上面的，是否运行Pod存在

1、污点等级

NoExecute 已经存在和创建的都移走
NoSchedule 创建的移走，已经存在的不影响
PreferNoSchedule 最后调度到这个节点上面

[root@master /]# kubectl explain node.spec.taints
#手动打上一个污点

[root@master qinhexing]# kubectl get pod -n dev -o wide
NAME            READY   STATUS    RESTARTS   AGE   IP               NODE    NOMINATED NODE   READINESS GATES
pod-first       1/1     Running   0          25m   10.244.104.17    node2   <none>           <none>
pod1-anti       1/1     Running   0          11m   10.244.166.145   node1   <none>           <none>
podaff-second   1/1     Running   0          25m   10.244.104.18    node2   <none>           <none>

[root@master qinhexing]# kubectl taint node node2 a=b:NoExecute
node/node2 tainted
[root@master qinhexing]# kubectl get pod -n dev 
NAME        READY   STATUS    RESTARTS   AGE
pod1-anti   1/1     Running   0          11m

2、容忍度

#查看node2上的污点
[root@master qinhexing]# kubectl describe node  node2 | grep -i taint
Taints:             a=b:NoExecute

#删除这个污点
[root@master qinhexing]# kubectl taint node node2 a:NoExecute-
node/node2 untainted

[root@master qinhexing]# kubectl describe node node2 | grep -i taint
Taints:             <none>

#容忍这个污点
equal等于的关系
exist就是key满足关系即可
[root@master qinhexing]# cat nodename.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: p1
  namespace: dev 
  labels:
    app: n1
spec:
  #nodeName: node2
  containers:
  - name: tomocat1
    image: docker.io/library/tomcat:8.5-jre8-alpine
    imagePullPolicy: IfNotPresent
    ports:
    - containerPort: 8080
  - name: busybox1
    image: docker.io/library/busybox:1.28
    imagePullPolicy: IfNotPresent
    command:
    - "/bin/sh"
    - "-c"
    - "sleep 3600000"
  tolerations:   #容忍度
  - key: a
    operator: "Equal"  #如果是Exists，value值为空即可
    value: b