【容器应用系列教程】Kubernetes调度系统

一、基于nodeName调度

指定在哪个node节点上，创建Pod

1.查看所有节点

[root@k8s-master ~]# kubectl get node
NAME                   STATUS   ROLES                  AGE   VERSION
k8s-master.linux.com   Ready    control-plane,master   8d    v1.20.7
k8s-node01.linux.com   Ready    <none>                 8d    v1.20.7
k8s-node02.linux.com   Ready    <none>                 8d    v1.20.7

2.基于`nodeName`调度

[root@k8s-master test]# vim test1.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test1-nginx
  namespace: default
spec:
  replicas: 2
  selector:
    matchLabels:
      app: test1-nginx
  template:
    metadata:
      labels:
        app: test1-nginx
    spec:
      nodeName: k8s-node02.linux.com	#指定所有Pod都运行在node02上
      containers:
      - name: test1-nginx
        image: nginx
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80

[root@k8s-master test]# kubectl create -f test1.yaml 
deployment.apps/test1-nginx created

3.查看`Pod`

可以看到所有Pod都跑在node02节点上

[root@k8s-master test]# kubectl get pods -o wide
NAME                           READY   STATUS    RESTARTS   AGE     IP                NODE                   NOMINATED NODE   READINESS GATES
test1-nginx-5bccbd59df-cfl6v   1/1     Running   0          2m29s   192.168.242.131   k8s-node02.linux.com   <none>           <none>
test1-nginx-5bccbd59df-r68jk   1/1     Running   0          2m29s   192.168.242.130   k8s-node02.linux.com   <none>           <none>

二、基于`nodeSelector`调度

1.查看所有节点的标签

[root@k8s-master test]# kubectl get node --show-labels
NAME                   STATUS   ROLES                  AGE   VERSION   LABELS
k8s-master.linux.com   Ready    control-plane,master   8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master.linux.com,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=
k8s-node01.linux.com   Ready    <none>                 8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01.linux.com,kubernetes.io/os=linux
k8s-node02.linux.com   Ready    <none>                 8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02.linux.com,kubernetes.io/os=linux

2.创建基于`nodeSelector`调度

[root@k8s-master test]# vim test2.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test2-nginx
  namespace: default
spec:
  replicas: 2
  selector:
    matchLabels:
      app: test2-nginx
  template:
    metadata:
      labels:
        app: test2-nginx
    spec:
      nodeSelector:	#nodeSelector调度类型
            name: wsjj	#在有name=wsjj的标签机器上创建
      containers:
      - name: test2-nginx
        image: nginx
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80

[root@k8s-master test]# kubectl create -f test2.yaml 
deployment.apps/test1-nginx created

3.查看`Pod`

可以看到所有Pod都是处于pending等待调度状态
这是因为，在所有节点上，都没有找到name=wsjj标签的机器

[root@k8s-master test]# kubectl get pod
NAME                           READY   STATUS    RESTARTS   AGE
test2-nginx-5f89c875cf-cfl6v   0/1     Pending   0          6s
test2-nginx-5f89c875cf-r68jk   0/1     Pending   0          6s

4.手动打标签

[root@k8s-master test]# kubectl label node k8s-node01.linux.com name=wsjj
node/k8s-node01.linux.com labeled

node01身上多了name=wsjj的标签

[root@k8s-master test]# kubectl get node --show-labels
NAME                   STATUS   ROLES                  AGE   VERSION   LABELS
k8s-master.linux.com   Ready    control-plane,master   8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master.linux.com,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=
k8s-node01.linux.com   Ready    <none>                 8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01.linux.com,kubernetes.io/os=linux,name=wsjj
k8s-node02.linux.com   Ready    <none>                 8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02.linux.com,kubernetes.io/os=linux

5.再次查看`Pod`

当有了标签后，Pod正常运行

[root@k8s-master test]# kubectl get pod -o wide
NAME                           READY   STATUS    RESTARTS   AGE    IP                NODE                   NOMINATED NODE   READINESS GATES
test2-nginx-5f89c875cf-cfl6v   1/1     Running   0          7m4s   192.168.201.194   k8s-node01.linux.com   <none>           <none>
test2-nginx-5f89c875cf-r68jk   1/1     Running   0          7m4s   192.168.201.195   k8s-node01.linux.com   <none>           <none>

6.删除标签

# kubectl label nodes <node_name> key-

[root@k8s-master test]# kubectl label nodes k8s-node01.linux.com name-
node/k8s-node01.linux.com labeled

可以看到即使标签被删除了，旧的`Pod`不会被影响

[root@k8s-master test]# kubectl get pod -o wide
NAME                           READY   STATUS    RESTARTS   AGE   IP                NODE                   NOMINATED NODE   READINESS GATES
test2-nginx-5f89c875cf-cfl6v   1/1     Running   0          11m   192.168.201.194   k8s-node01.linux.com   <none>           <none>
test2-nginx-5f89c875cf-r68jk   1/1     Running   0          11m   192.168.201.195   k8s-node01.linux.com   <none>           <none>

[root@k8s-master test]# kubectl delete pod test2-nginx-5f89c875cf-cfl6v
pod "test2-nginx-5f89c875cf-cfl6v" deleted

[root@k8s-master test]# kubectl get pod -o wide
NAME                           READY   STATUS    RESTARTS   AGE   IP                NODE                   NOMINATED NODE   READINESS GATES
test2-nginx-5f89c875cf-7d9tj   0/1     Pending   0          17s   <none>            <none>                 <none>           <none>
test2-nginx-5f89c875cf-r68jk   1/1     Running   0          12m   192.168.201.195   k8s-node01.linux.com   <none>           <none>

7.修改标签的值

# kubectl label nodes <node_name> key=value --overwrite

三、基于`nodeAffinity`节点亲缘性

使用nodeAffinity可以在部署Pod时，将Pod调度运行到符合特定条件或不符合特定条件的节点

1.`nodeAffinity`调度策略

`requiredDuringSchedulingIgnoredDuringExecution`

表示pod必须部署到满足条件的节点上，如果没有满足条件的节点，就不停重试。
IgnoreDuringExecution表示pod部署之后运行的时候，如果节点标签发生了变化，不再满足pod指定的条件，pod也会继续运行。

`requiredDuringSchedulingRequiredDuringExecution`

表示pod必须部署到满足条件的节点上，如果没有满足条件的节点，就不停重试
RequiredDuringExecution表示pod部署之后运行的时候，如果节点标签发生了变化，不再满足pod指定的条件，则重新选择符合要求的节点

`preferredDuringSchedulingIgnoredDuringExecution`

表示优先部署到满足条件的节点上，如果没有满足条件的节点，就忽略这些条件，按照正常逻辑部署

`preferredDuringSchedulingRequiredDuringExecution`

表示优先部署到满足条件的节点上，如果没有满足条件的节点，就忽略这些条件，按照正常逻辑部署。
RequiredDuringExecution表示如果后面节点标签发生了变化，满足了条件，则重新调度到满足条件的节点。

2.案例1：基于`nodeAffinity`节点亲缘性调度

为`node02`打标记

[root@k8s-master test]# kubectl label node k8s-node02.linux.com yewu=dev
node/k8s-node02.linux.com labeled

[root@k8s-master test]# kubectl get node --show-labels
NAME                   STATUS   ROLES                  AGE   VERSION   LABELS
k8s-master.linux.com   Ready    control-plane,master   8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master.linux.com,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=
k8s-node01.linux.com   Ready    <none>                 8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01.linux.com,kubernetes.io/os=linux
k8s-node02.linux.com   Ready    <none>                 8d    v1.20.7   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02.linux.com,kubernetes.io/os=linux,yewu=dev

创建`Deployment`测试

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test3-nginx
  namespace: default
  labels:
    app: test3-nginx
spec:
  replicas: 2
  selector:
    matchLabels:
      app: test3-nginx
  template:
    metadata:
      labels:
        app: test3-nginx
    spec:
      affinity:
        nodeAffinity:	#类型
          requiredDuringSchedulingIgnoredDuringExecution:	#指定策略
            nodeSelectorTerms:
            - matchExpressions:
              - key: yewu	#键
                operator: NotIn	#表示label的值不在某个列表中，类似取反
                values:
                - dev	#值
      containers:
      - name: test3-nginx
        image: nginx
        ports:
        - containerPort: 80

支持的操作符：
In: label的值在某个列表中
NotIn：label的值不在某个列表中
Exists：某个label存在
DoesNotExist：某个label不存在
Gt：label的值大于某个值（字符串比较）
Lt：label的值小于某个值（字符串比较）

[root@k8s-master test]# kubectl create -f test3.yml
deployment.apps/test3-nginx created

查看结果

由于取反的原因，node02身上有符合要求的标签，所以取反到node01身上创建

[root@k8s-master test]# kubectl get pod -o wide
NAME                           READY   STATUS    RESTARTS   AGE     IP                NODE                   NOMINATED NODE   READINESS GATES
test3-nginx-654cf47554-bzbh4   1/1     Running   0          2m34s   192.168.201.195   k8s-node01.linux.com   <none>           <none>
test3-nginx-654cf47554-cwggc   1/1     Running   0          2m34s   192.168.201.196   k8s-node01.linux.com   <none>           <none>

3.案例2：基于`nodeAffinity`节点亲缘性调度

[root@k8s-master test]# vim test4.yml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test4-nginx
  namespace: default
  labels:
    app: test4-nginx
spec:
  replicas: 2
  selector:
    matchLabels:
      app: test4-nginx
  template:
    metadata:
      labels:
        app: test4-nginx
    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                - key: kubernetes.io/os
                  operator: In
                  values:
                    - linux
          preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 1
              preference:
                matchExpressions:
                  - key: yewu
                    operator: In
                    values:
                     - dev
      containers:
      - name: test4-nginx
        image: nginx
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80

[root@k8s-master test]# kubectl create -f test4.yml
deployment.apps/test4-nginx created

查看`Pod`

本案例中同时设置了requiredDuringSchedulingIgnoredDuringExecution和preferredDuringSchedulingIgnoredDuringExecution两种策略，表示的意思为将Pod要调度到具有kubernetes.io/os=linux这个标签的节点时，在满足该条件的情况下，preferredDuringSchedulingIgnoredDuringExecution表示的意思是优先将其调度到同时具有yewu=dev的节点上

[root@k8s-master test]# kubectl get pods -o wide
NAME                           READY   STATUS    RESTARTS   AGE    IP                NODE                   NOMINATED NODE   READINESS GATES
test4-nginx-5bf9894f8c-87wsf   1/1     Running   0          2m3s   192.168.242.132   k8s-node02.linux.com   <none>           <none>
test4-nginx-5bf9894f8c-nznnn   1/1     Running   0          2m3s   192.168.242.131   k8s-node02.linux.com   <none>           <none>

四、污点`Taint`、容忍`Toleration`

Taint和toleration相互配合，可以用来避免pod被分配到不合适的节点上。每个节点上都可以应用一个或多个taint，这表示对于那些不能容忍这些taint的pod，是不会被该节点接受的。如果将toleration应用于pod上，则表示这些pod可以被调度到具有相应taint的节点上。

1.污点`Taint`管理操作

查看`Taint`污点

由于我是基于kubeadm部署的集群，集群会默认为Master节点配置一个污点，防止资源被调度到Master节点上。

[root@k8s-master test]# kubectl describe node k8s-master.linux.com | grep -i Taints
Taints:             node-role.kubernetes.io/master:NoSchedule

为节点添加`Taint`污点

# kubectl taint nodes <node_name> key=value:{NoSchedule|NoExecute|PreferNoSchedule}

污点的影响策略:
NoSchedule
- 新的不能容忍该污点的Pod不会再调度过来
- 但已经运行在该节点上的Pod不会受影响
NoExecute
- 新的不能容忍该污点的Pod不会再调度过来
- 同时该节点上的所有Pod都会被驱逐
PreferNoSchedule
- 尽量不向该节点调度

删除`Taint`污点

# kubectl taint node <node_name> key:{NoSchedule|NoExecute|PreferNoSchedule}-

3.示例：为`node02`添加`Tatin`污点，指定策略为`NoExecute`

[root@k8s-master ~]# kubectl taint nodes k8s-node02.linux.com AA=10:NoExecute

可以看到`node02`身上的所有`Pod`都被驱逐了

[root@k8s-master ~]# kubectl get pods -o wide
NAME                           READY   STATUS              RESTARTS   AGE   IP                NODE                   NOMINATED NODE   READINESS GATES
test-client                    1/1     Running             4          10d   10.88.0.9         k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-6x8mz   0/1     ContainerCreating   0          3s    <none>            k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-7wkhg   0/1     ContainerCreating   0          3s    <none>            k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-c8pf9   1/1     Running             0          10m   172.168.201.231   k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-fbfxp   1/1     Running             0          10m   172.168.201.230   k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-j4j62   1/1     Terminating         0          10m   172.168.242.138   k8s-node02.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-j4zkd   0/1     ContainerCreating   0          3s    <none>            k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-lfct5   1/1     Terminating         0          10m   172.168.242.139   k8s-node02.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-nnlml   1/1     Terminating         0          10m   172.168.242.135   k8s-node02.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-p4nrz   0/1     ContainerCreating   0          3s    <none>            k8s-node01.linux.com   <none>           <none>
test1-nginx-6fcf9564d5-t8jdc   1/1     Terminating         0          10m   172.168.242.134   k8s-node02.linux.com   <none>           <none>

删除污点

# kubectl taint node 节点名 键-

[root@k8s-master ~]# kubectl taint nodes k8s-node02.linux.com AA-

4.容忍`Toleration`

一般容忍Toleration和污点Taint是配合使用的
当用户想要在，存在Taint污点的Node节点上创建Pod的时候，就需要有容忍Toleration的存在

添加`Tatin`污点

[root@k8s-master ~]# kubectl taint nodes k8s-node02.linux.com AA=10:NoExecute

创建`Pod`配置容忍`Toleration`策略

[root@k8s-master test]# vim test5.yml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: test5-nginx
  namespace: default
  labels:
    app: test5-nginx
spec:
  replicas: 2
  selector:
    matchLabels:
      app: test5-nginx
  template:
    metadata:
      labels:
        app: test5-nginx
    spec:
      containers:
      - name: test5-nginx
        image: nginx
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80

      tolerations:
        - key: "AA"
          operator: "Equal"   
          value: "10"
          effect: NoExecute
          tolerationSeconds: 30	#容忍时间可以不写

[root@k8s-master test]# kubectl create -f test5.yml

配置项说明：
tolerations添加容忍策略
key指定所容忍的key，要与对应污点的key一致
value指定所容忍的value，要与对应污点的value一致
effect指定所容忍的策略，一定要与污点指定的策略一致
tolerationSeconds指定容忍的超时时间，到达该时间后，Pod将会被驱逐

容忍的匹配规则也可以写多个

tolerations:
- key: "key1"
  operator: "Equal"
  value: "value1"
  effect: "NoSchedule"
- key: "key1"
  operator: "Equal"
  value: "value1"
  effect: "NoExecute"

查看`Pod`

[root@k8s-master ]# kubectl get pods -o wide

NAME                           READY   STATUS    RESTARTS   AGE   IP                NODE                   NOMINATED NODE   READINESS GATES
test1-nginx-654d4cc68c-2mh68   1/1     Running   0          6s    172.168.242.149   k8s-node02.linux.com   <none>           <none>
test1-nginx-654d4cc68c-8z8kw   1/1     Running   0          6s    172.168.242.147   k8s-node02.linux.com   <none>           <none>

删除污点`Taint`

[root@k8s-master ~]# kubectl taint node k8s-node02.linux.com AA-

如果觉得文章对你有用，请随意赞赏

k8s kubernetes centos7 容器容器编排工具调度

【容器应用系列教程】Kubernetes调度系统

https://www.wsjj.top/archives/148

作者

网笙久久

发布于

2023-06-16

更新于

2023-07-19

许可协议

CC BY 4.0

【容器应用系列教程】Kubernetes调度系统

【容器应用系列教程】Kubernetes调度系统

一、基于nodeName调度

1.查看所有节点

2.基于nodeName调度

3.查看Pod

二、基于nodeSelector调度

1.查看所有节点的标签

2.创建基于nodeSelector调度

3.查看Pod

4.手动打标签

5.再次查看Pod

6.删除标签

可以看到即使标签被删除了，旧的Pod不会被影响

7.修改标签的值

三、基于nodeAffinity节点亲缘性

1.nodeAffinity调度策略

requiredDuringSchedulingIgnoredDuringExecution

requiredDuringSchedulingRequiredDuringExecution

preferredDuringSchedulingIgnoredDuringExecution

preferredDuringSchedulingRequiredDuringExecution

2.案例1：基于nodeAffinity节点亲缘性调度

为node02打标记

创建Deployment测试

查看结果

3.案例2：基于nodeAffinity节点亲缘性调度

查看Pod

四、污点Taint、容忍Toleration

1.污点Taint管理操作

查看Taint污点

为节点添加Taint污点

删除Taint污点

3.示例：为node02添加Tatin污点，指定策略为NoExecute

可以看到node02身上的所有Pod都被驱逐了

删除污点

4.容忍Toleration

添加Tatin污点

创建Pod配置容忍Toleration策略

容忍的匹配规则也可以写多个

查看Pod

删除污点Taint

作者

发布于

更新于

许可协议

评论

2.基于`nodeName`调度

3.查看`Pod`

二、基于`nodeSelector`调度

2.创建基于`nodeSelector`调度

3.查看`Pod`

5.再次查看`Pod`

可以看到即使标签被删除了，旧的`Pod`不会被影响

三、基于`nodeAffinity`节点亲缘性

1.`nodeAffinity`调度策略

`requiredDuringSchedulingIgnoredDuringExecution`

`requiredDuringSchedulingRequiredDuringExecution`

`preferredDuringSchedulingIgnoredDuringExecution`

`preferredDuringSchedulingRequiredDuringExecution`

2.案例1：基于`nodeAffinity`节点亲缘性调度

为`node02`打标记

创建`Deployment`测试

3.案例2：基于`nodeAffinity`节点亲缘性调度

查看`Pod`

四、污点`Taint`、容忍`Toleration`

1.污点`Taint`管理操作

查看`Taint`污点

为节点添加`Taint`污点

删除`Taint`污点

3.示例：为`node02`添加`Tatin`污点，指定策略为`NoExecute`

可以看到`node02`身上的所有`Pod`都被驱逐了

4.容忍`Toleration`

添加`Tatin`污点

创建`Pod`配置容忍`Toleration`策略

查看`Pod`

删除污点`Taint`