污点和污点容忍

leo_mmmm

已于 2024-07-28 21:35:36 修改

阅读量326

点赞数 3

分类专栏： k8s 文章标签： kubernetes 容器

于 2024-07-28 21:31:34 首次发布

本文链接：https://blog.csdn.net/weixin_46466657/article/details/140756009

版权

k8s 专栏收录该内容

13 篇文章 1 订阅

订阅专栏

1.概述

        污点就是给k8s集群中的节点设置的污点，通过设置污点，来规划资源创建时所在的节点

污点的类型：

        PreferNoshedule：节点设置这个污点类型后表示该节点接受调度，但是会降低调度的概率

        NoShedule：表示该节点不接受新的调度，以前有的资源，可以继续存在

        NoExecute：表示不接受新的调度，同时会驱逐以前的资源调度

2.污点的管理

·查看污点

[root@master job]# kubectl describe nodes |grep -i taint
Taints:             node-role.kubernetes.io/master:NoSchedule
Taints:             <none>
Taints:             <none>

查看前两行

[root@master job]# kubectl describe nodes |grep -i taint -A 2
Taints:             node-role.kubernetes.io/master:NoSchedule
Unschedulable:      false
Lease:
--
Taints:             <none>
Unschedulable:      false
Lease:
--
Taints:             <none>
Unschedulable:      false
Lease:

·创建污点

1.创建前查看pod详情

[root@master job]# kubectl get pods -o wide
NAME                     READY   STATUS              RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
dm002-7b46896f67-7q5vd   1/1     Running             0          14s   10.100.2.158   worker2   <none>           <none>
dm002-7b46896f67-ckq86   1/1     Running             0          14s   10.100.2.155   worker2   <none>           <none>
dm002-7b46896f67-nspxv   1/1     Running             0          14s   10.100.2.156   worker2   <none>           <none>
dm002-7b46896f67-r92tj   1/1     Running             0          14s   10.100.2.154   worker2   <none>           <none>
dm002-7b46896f67-wmggx   1/1     Running             0          13s   10.100.2.157   worker2   <none>           <none>

2.创建污点

创建污点有两种方式：

第一种：key=value:污点类型

第二种：key:污点类型

3.创建污点

[root@master job]# kubectl taint node worker2 k8s=demo:NoExecute
node/worker2 tainted

4.创建污点后，查看pod

[root@master job]# kubectl get pods -o wide
NAME                     READY   STATUS              RESTARTS   AGE   IP       NODE      NOMINATED NODE   READINESS GATES
dm002-7b46896f67-2d64t   0/1     Pending             0          3s    <none>   <none>    <none>           <none>
dm002-7b46896f67-2tczc   0/1     Pending             0          3s    <none>   <none>    <none>           <none>
dm002-7b46896f67-6lbmn   0/1     Pending             0          4s    <none>   <none>    <none>           <none>
dm002-7b46896f67-8xt9t   0/1     Pending             0          4s    <none>   <none>    <none>           <none>
dm002-7b46896f67-w8mjs   0/1     Pending             0          3s    <none>   <none>    <none>           <none>

·删除污点

[root@master job]# kubectl taint node worker2 k8s=demo:NoExecute-
node/worker2 untainted

[root@master job]# kubectl get pods -o wide
NAME                     READY   STATUS              RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
dm002-7b46896f67-2s8vc   1/1     Running             0          30s   10.100.2.177   worker2   <none>           <none>
dm002-7b46896f67-phzcb   1/1     Running             0          30s   10.100.2.178   worker2   <none>           <none>
dm002-7b46896f67-q8z6h   1/1     Running             0          30s   10.100.2.174   worker2   <none>           <none>
dm002-7b46896f67-x5dzv   1/1     Running             0          30s   10.100.2.175   worker2   <none>           <none>
dm002-7b46896f67-xwvg5   1/1     Running             0          30s   10.100.2.176   worker2   <none>           <none>

·修改污点

[root@master job]# kubectl taint node worker2 k8s=test:NoExecute
node/worker2 tainted
[root@master job]# kubectl taint node worker2 k8s=demo:NoSchedule --overwrite
node/worker2 modified

3.污点容忍tolerations

通过给k8s集群节点设置不同类型的污点，来控制资源创建的节点范围

污点容忍就是即便节点设置了污点，资源也可以创建在这个节点上；

即在节点上设置了污点之后，在创建资源时，可以无视这个节点上的污点规则，也就可以在有污点的节点上创建资源

假设一个节点上有两个污点，但是你创建pod的时候，还想要在这个节点上创建，那么你就需要再资源清单中，写入容忍这两个污点，才会创建成功

·示例

利用之前给k8s233创建两个污点

1.编辑一个正常的资源清单

[root@master job]# cat deploy.yaml 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: test001
spec:
  replicas: 5
  selector:
    matchLabels:
      k8s: k8s
  template:
    metadata:
      labels:
        k8s: k8s
    spec:
      containers:
      - name: test-c
        image: nginx:1.20.1-alpine
        ports:
        - name: port
          containerPort: 80

创建查看资源，由于设置了污点，所以节点都创建在了worker1节点上

[root@master job]# kubectl apply -f deploy.yaml 
deployment.apps/test001 created
[root@master job]# kubectl get pods -o wide
NAME                      READY   STATUS              RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
test001-f8998c97c-5hwkr   1/1     Running             0          62s   10.100.1.227   worker1   <none>           <none>
test001-f8998c97c-hspvt   1/1     Running             0          62s   10.100.1.229   worker1   <none>           <none>
test001-f8998c97c-mrxvg   1/1     Running             0          62s   10.100.1.226   worker1   <none>           <none>
test001-f8998c97c-q7lhx   1/1     Running             0          62s   10.100.1.225   worker1   <none>           <none>
test001-f8998c97c-v85ws   1/1     Running             0          62s   10.100.1.228   worker1   <none>           <none>

2. 修改资源清单，设置污点容忍

[root@master job]# cat deploy.yaml 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: test001
spec:
  replicas: 5
  selector:
    matchLabels:
      k8s: k8s
  template:
    metadata:
      labels:
        k8s: k8s
    spec:
      #设置污点容忍
      tolerations:
      #指定污点的key
      - key: node-role.kubernets.io/master
        #指定污点的类型
        effect: NoSchedule
        #设置key和value的关系
        #--Exists：key匹配所有value
        #--Equsl：key和value必须全部匹配到
        operator: Exists
      - key: k8s
        value: test
        effect: NoExecute
        operator: Equal
      - key: k8s
        value: demo
        effect: NoSchedule
        operator: Equal
      containers:
      - name: test-c
        image: nginx:1.20.1-alpine
        ports:
        - name: port
          containerPort: 80

重新创建资源

[root@master job]# kubectl apply -f deploy.yaml 
deployment.apps/test001 configured

查看pod详情，发现worker1和worker2节点上都有

[root@master job]# kubectl get pods -o wide
NAME                       READY   STATUS              RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
test001-7955599844-2lt2x   1/1     Running             0          30s   10.100.2.187   worker2   <none>           <none>
test001-7955599844-b56p8   1/1     Running             0          30s   10.100.1.250   worker1   <none>           <none>
test001-7955599844-j47bv   1/1     Running             0          30s   10.100.2.188   worker2   <none>           <none>
test001-7955599844-ncrqf   1/1     Running             0          30s   10.100.1.249   worker1   <none>           <none>
test001-7955599844-zbpmn   1/1     Running             0          30s   10.100.2.189   worker2   <none>           <none>

3.修改资源清单，无视所有污点

[root@master job]# cat deploy.yaml 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: test001
spec:
  replicas: 5
  selector:
    matchLabels:
      k8s: k8s
  template:
    metadata:
      labels:
        k8s: k8s
    spec:
      tolerations:
      #不写污点的key的相关属性表示匹配所有key
      - operator: Exists
      containers:
      - name: test-c
        image: nginx:1.20.1-alpine
        ports:
        - name: port
          containerPort: 80

重新创建资源

[root@master job]# kubectl apply -f deploy.yaml 
deployment.apps/test001 configured

查看pod详情

[root@master job]# kubectl get pods -o wide
NAME                       READY   STATUS              RESTARTS   AGE     IP             NODE      NOMINATED NODE   READINESS GATES
test001-6588d66c78-kmzj8   1/1     Running             0          14s     10.100.2.195   worker2   <none>           <none>
test001-6588d66c78-rz9rl   1/1     Running             0          14s     10.100.1.10    worker1   <none>           <none>
test001-6588d66c78-t4t5f   1/1     Running             0          14s     10.100.0.7     master    <none>           <none>
test001-6588d66c78-v9pvp   1/1     Running             0          14s     10.100.2.196   worker2   <none>           <none>
test001-6588d66c78-xssvd   1/1     Running             0          15s     10.100.1.8     worker1   <none>           <none>