Kubernetes K8S之Taints污点与Tolerations容忍详解

 

Kubernetes K8S之Taints污点与Tolerations容忍详解与示例

 

主机配置规划

服务器名称(hostname) 系统版本 配置 内网IP 外网IP(模拟)
k8s-master CentOS7.7 2C/4G/20G 172.16.1.110 10.0.0.110
k8s-node01 CentOS7.7 2C/4G/20G 172.16.1.111 10.0.0.111
k8s-node02 CentOS7.7 2C/4G/20G 172.16.1.112 10.0.0.112

 

Taints污点和Tolerations容忍概述

节点和Pod亲和力,是将Pod吸引到一组节点【根据拓扑域】(作为优选或硬性要求)。污点(Taints)则相反,它们允许一个节点排斥一组Pod。

容忍(Tolerations)应用于pod,允许(但不强制要求)pod调度到具有匹配污点的节点上。

污点(Taints)和容忍(Tolerations)共同作用,确保pods不会被调度到不适当的节点。一个或多个污点应用于节点;这标志着该节点不应该接受任何不容忍污点的Pod。

说明:我们在平常使用中发现pod不会调度到k8s的master节点,就是因为master节点存在污点。

 

Taints污点

Taints污点的组成

使用kubectl taint命令可以给某个Node节点设置污点,Node被设置污点之后就和Pod之间存在一种相斥的关系,可以让Node拒绝Pod的调度执行,甚至将Node上已经存在的Pod驱逐出去。

每个污点的组成如下:

key=value:effect

每个污点有一个key和value作为污点的标签,effect描述污点的作用。当前taint effect支持如下选项

  • NoSchedule:表示K8S将不会把Pod调度到具有该污点的Node节点上
  • PreferNoSchedule:表示K8S将尽量避免把Pod调度到具有该污点的Node节点上
  • NoExecute:表示K8S将不会把Pod调度到具有该污点的Node节点上,同时会将Node上已经存在的Pod驱逐出去

 

污点taint的NoExecute详解

taint 的 effect 值 NoExecute,它会影响已经在节点上运行的 pod:

  • 如果 pod 不能容忍 effect 值为 NoExecute 的 taint,那么 pod 将马上被驱逐
  • 如果 pod 能够容忍 effect 值为 NoExecute 的 taint,且在 toleration 定义中没有指定 tolerationSeconds,则 pod 会一直在这个节点上运行。
  • 如果 pod 能够容忍 effect 值为 NoExecute 的 taint,但是在toleration定义中指定了 tolerationSeconds,则表示 pod 还能在这个节点上继续运行的时间长度。

 

Taints污点设置

污点(Taints)查看

k8s master节点查看

kubectl describe node k8s-master

 

k8s node查看

kubectl describe node k8s-node01

 

污点(Taints)添加

 1 [root@k8s-master taint]# kubectl taint nodes k8s-node01 check=zhang:NoSchedule
 2 node/k8s-node01 tainted
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl describe node k8s-node01
 5 Name:               k8s-node01
 6 Roles:              <none>
 7 Labels:             beta.kubernetes.io/arch=amd64
 8                     beta.kubernetes.io/os=linux
 9                     cpu-num=12
10                     disk-type=ssd
11                     kubernetes.io/arch=amd64
12                     kubernetes.io/hostname=k8s-node01
13                     kubernetes.io/os=linux
14                     mem-num=48
15 Annotations:        flannel.alpha.coreos.com/backend-data: {"VtepMAC":"3e:15:bb:f8:85:dc"}
16                     flannel.alpha.coreos.com/backend-type: vxlan
17                     flannel.alpha.coreos.com/kube-subnet-manager: true
18                     flannel.alpha.coreos.com/public-ip: 10.0.0.111
19                     kubeadm.alpha.kubernetes.io/cri-socket: /var/run/dockershim.sock
20                     node.alpha.kubernetes.io/ttl: 0
21                     volumes.kubernetes.io/controller-managed-attach-detach: true
22 CreationTimestamp:  Tue, 12 May 2020 16:50:54 +0800
23 Taints:             check=zhang:NoSchedule   ### 可见已添加污点
24 Unschedulable:      false

在k8s-node01节点添加了一个污点(taint),污点的key为check,value为zhang,污点effect为NoSchedule。这意味着没有pod可以调度到k8s-node01节点,除非具有相匹配的容忍。

 

污点(Taints)删除

 1 [root@k8s-master taint]# kubectl taint nodes k8s-node01 check:NoExecute-
 2 ##### 或者
 3 [root@k8s-master taint]# kubectl taint nodes k8s-node01 check=zhang:NoSchedule-
 4 node/k8s-node01 untainted
 5 [root@k8s-master taint]# 
 6 [root@k8s-master taint]# kubectl describe node k8s-node01
 7 Name:               k8s-node01
 8 Roles:              <none>
 9 Labels:             beta.kubernetes.io/arch=amd64
10                     beta.kubernetes.io/os=linux
11                     cpu-num=12
12                     disk-type=ssd
13                     kubernetes.io/arch=amd64
14                     kubernetes.io/hostname=k8s-node01
15                     kubernetes.io/os=linux
16                     mem-num=48
17 Annotations:        flannel.alpha.coreos.com/backend-data: {"VtepMAC":"3e:15:bb:f8:85:dc"}
18                     flannel.alpha.coreos.com/backend-type: vxlan
19                     flannel.alpha.coreos.com/kube-subnet-manager: true
20                     flannel.alpha.coreos.com/public-ip: 10.0.0.111
21                     kubeadm.alpha.kubernetes.io/cri-socket: /var/run/dockershim.sock
22                     node.alpha.kubernetes.io/ttl: 0
23                     volumes.kubernetes.io/controller-managed-attach-detach: true
24 CreationTimestamp:  Tue, 12 May 2020 16:50:54 +0800
25 Taints:             <none>   ### 可见已删除污点
26 Unschedulable:      false

 

Tolerations容忍

设置了污点的Node将根据taint的effect:NoSchedule、PreferNoSchedule、NoExecute和Pod之间产生互斥的关系,Pod将在一定程度上不会被调度到Node上。

但我们可以在Pod上设置容忍(Tolerations),意思是设置了容忍的Pod将可以容忍污点的存在,可以被调度到存在污点的Node上。

pod.spec.tolerations示例

 1 tolerations:
 2 - key: "key"
 3   operator: "Equal"
 4   value: "value"
 5   effect: "NoSchedule"
 6 ---
 7 tolerations:
 8 - key: "key"
 9   operator: "Exists"
10   effect: "NoSchedule"
11 ---
12 tolerations:
13 - key: "key"
14   operator: "Equal"
15   value: "value"
16   effect: "NoExecute"
17   tolerationSeconds: 3600

重要说明:

  • 其中key、value、effect要与Node上设置的taint保持一致
  • operator的值为Exists时,将会忽略value;只要有key和effect就行
  • tolerationSeconds:表示pod 能够容忍 effect 值为 NoExecute 的 taint;当指定了 tolerationSeconds【容忍时间】,则表示 pod 还能在这个节点上继续运行的时间长度。

 

当不指定key值时

当不指定key值和effect值时,且operator为Exists,表示容忍所有的污点【能匹配污点所有的keys,values和effects】

1 tolerations:
2 - operator: "Exists"

 

当不指定effect值时

当不指定effect值时,则能匹配污点key对应的所有effects情况

1 tolerations:
2 - key: "key"
3   operator: "Exists"

 

当有多个Master存在时

当有多个Master存在时,为了防止资源浪费,可以进行如下设置:

1 kubectl taint nodes Node-name node-role.kubernetes.io/master=:PreferNoSchedule

 

多个Taints污点和多个Tolerations容忍怎么判断

可以在同一个node节点上设置多个污点(Taints),在同一个pod上设置多个容忍(Tolerations)。Kubernetes处理多个污点和容忍的方式就像一个过滤器:从节点的所有污点开始,然后忽略可以被Pod容忍匹配的污点;保留其余不可忽略的污点,污点的effect对Pod具有显示效果:特别是:

  • 如果有至少一个不可忽略污点,effect为NoSchedule,那么Kubernetes将不调度Pod到该节点
  • 如果没有effect为NoSchedule的不可忽视污点,但有至少一个不可忽视污点,effect为PreferNoSchedule,那么Kubernetes将尽量不调度Pod到该节点
  • 如果有至少一个不可忽视污点,effect为NoExecute,那么Pod将被从该节点驱逐(如果Pod已经在该节点运行),并且不会被调度到该节点(如果Pod还未在该节点运行)

 

污点和容忍示例

Node污点为NoExecute的示例

记得把已有的污点清除,以免影响测验。

节点上的污点设置(Taints)

实现如下污点

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:
3 k8s-node02 污点为:

 

污点添加操作如下:
“无,本次无污点操作”

污点查看操作如下:

1 kubectl describe node k8s-master | grep 'Taints' -A 5
2 kubectl describe node k8s-node01 | grep 'Taints' -A 5
3 kubectl describe node k8s-node02 | grep 'Taints' -A 5

除了k8s-master默认的污点,在k8s-node01、k8s-node02无污点。

 

污点为NoExecute示例

yaml文件

 1 [root@k8s-master taint]# pwd
 2 /root/k8s_practice/scheduler/taint
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# cat noexecute_tolerations.yaml 
 5 apiVersion: apps/v1
 6 kind: Deployment
 7 metadata:
 8   name: noexec-tolerations-deploy
 9   labels:
10     app: noexectolerations-deploy
11 spec:
12   replicas: 6
13   selector:
14     matchLabels:
15       app: myapp
16   template:
17     metadata:
18       labels:
19         app: myapp
20     spec:
21       containers:
22       - name: myapp-pod
23         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1
24         imagePullPolicy: IfNotPresent
25         ports:
26           - containerPort: 80
27        # 有容忍并有 tolerationSeconds 时的格式
28 #      tolerations:
29 #      - key: "check-mem"
30 #        operator: "Equal"
31 #        value: "memdb"
32 #        effect: "NoExecute"
33 #        # 当Pod将被驱逐时,Pod还可以在Node节点上继续保留运行的时间
34 #        tolerationSeconds: 30

 

运行yaml文件

 1 [root@k8s-master taint]# kubectl apply -f noexecute_tolerations.yaml 
 2 deployment.apps/noexec-tolerations-deploy created
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl get deploy -o wide
 5 NAME                        READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR
 6 noexec-tolerations-deploy   6/6     6            6           10s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp
 7 [root@k8s-master taint]# 
 8 [root@k8s-master taint]# kubectl get pod -o wide
 9 NAME                                         READY   STATUS    RESTARTS   AGE   IP             NODE         NOMINATED NODE   READINESS GATES
10 noexec-tolerations-deploy-85587896f9-2j848   1/1     Running   0          15s   10.244.4.101   k8s-node01   <none>           <none>
11 noexec-tolerations-deploy-85587896f9-jgqkn   1/1     Running   0          15s   10.244.2.141   k8s-node02   <none>           <none>
12 noexec-tolerations-deploy-85587896f9-jmw5w   1/1     Running   0          15s   10.244.2.142   k8s-node02   <none>           <none>
13 noexec-tolerations-deploy-85587896f9-s8x95   1/1     Running   0          15s   10.244.4.102   k8s-node01   <none>           <none>
14 noexec-tolerations-deploy-85587896f9-t82fj   1/1     Running   0          15s   10.244.4.103   k8s-node01   <none>           <none>
15 noexec-tolerations-deploy-85587896f9-wx9pz   1/1     Running   0          15s   10.244.2.143   k8s-node02   <none>           <none>

由上可见,pod是在k8s-node01、k8s-node02平均分布的。

 

添加effect为NoExecute的污点
kubectl taint nodes k8s-node02 check-mem=memdb:NoExecute

 

此时所有节点污点为

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:
3 k8s-node02 污点为:check-mem=memdb:NoExecute

 

之后再次查看pod信息

1 [root@k8s-master taint]# kubectl get pod -o wide
2 NAME                                         READY   STATUS    RESTARTS   AGE    IP             NODE         NOMINATED NODE   READINESS GATES
3 noexec-tolerations-deploy-85587896f9-2j848   1/1     Running   0          2m2s   10.244.4.101   k8s-node01   <none>           <none>
4 noexec-tolerations-deploy-85587896f9-ch96j   1/1     Running   0          8s     10.244.4.106   k8s-node01   <none>           <none>
5 noexec-tolerations-deploy-85587896f9-cjrkb   1/1     Running   0          8s     10.244.4.105   k8s-node01   <none>           <none>
6 noexec-tolerations-deploy-85587896f9-qbq6d   1/1     Running   0          7s     10.244.4.104   k8s-node01   <none>           <none>
7 noexec-tolerations-deploy-85587896f9-s8x95   1/1     Running   0          2m2s   10.244.4.102   k8s-node01   <none>           <none>
8 noexec-tolerations-deploy-85587896f9-t82fj   1/1     Running   0          2m2s   10.244.4.103   k8s-node01   <none>           <none>

由上可见,在k8s-node02节点上的pod已被驱逐,驱逐的pod被调度到了k8s-node01节点。

 

Pod没有容忍时(Tolerations)

记得把已有的污点清除,以免影响测验。

节点上的污点设置(Taints)

实现如下污点

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:check-nginx=web:PreferNoSchedule
3 k8s-node02 污点为:check-nginx=web:NoSchedule

 

污点添加操作如下:

1 kubectl taint nodes k8s-node01 check-nginx=web:PreferNoSchedule
2 kubectl taint nodes k8s-node02 check-nginx=web:NoSchedule

 

污点查看操作如下:

1 kubectl describe node k8s-master | grep 'Taints' -A 5
2 kubectl describe node k8s-node01 | grep 'Taints' -A 5
3 kubectl describe node k8s-node02 | grep 'Taints' -A 5

 

无容忍示例

yaml文件

 1 [root@k8s-master taint]# pwd
 2 /root/k8s_practice/scheduler/taint
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# cat no_tolerations.yaml 
 5 apiVersion: apps/v1
 6 kind: Deployment
 7 metadata:
 8   name: no-tolerations-deploy
 9   labels:
10     app: notolerations-deploy
11 spec:
12   replicas: 5
13   selector:
14     matchLabels:
15       app: myapp
16   template:
17     metadata:
18       labels:
19         app: myapp
20     spec:
21       containers:
22       - name: myapp-pod
23         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1
24         imagePullPolicy: IfNotPresent
25         ports:
26           - containerPort: 80

 

运行yaml文件

 1 [root@k8s-master taint]# kubectl apply -f no_tolerations.yaml 
 2 deployment.apps/no-tolerations-deploy created
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl get deploy -o wide
 5 NAME                    READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR
 6 no-tolerations-deploy   5/5     5            5           9s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp
 7 [root@k8s-master taint]# 
 8 [root@k8s-master taint]# kubectl get pod -o wide
 9 NAME                                     READY   STATUS    RESTARTS   AGE   IP            NODE         NOMINATED NODE   READINESS GATES
10 no-tolerations-deploy-85587896f9-6bjv8   1/1     Running   0          16s   10.244.4.54   k8s-node01   <none>           <none>
11 no-tolerations-deploy-85587896f9-hbbjb   1/1     Running   0          16s   10.244.4.58   k8s-node01   <none>           <none>
12 no-tolerations-deploy-85587896f9-jlmzw   1/1     Running   0          16s   10.244.4.56   k8s-node01   <none>           <none>
13 no-tolerations-deploy-85587896f9-kfh2c   1/1     Running   0          16s   10.244.4.55   k8s-node01   <none>           <none>
14 no-tolerations-deploy-85587896f9-wmp8b   1/1     Running   0          16s   10.244.4.57   k8s-node01   <none>           <none>

由上可见,因为k8s-node02节点的污点check-nginx 的effect为NoSchedule,说明pod不能被调度到该节点。此时k8s-node01节点的污点check-nginx 的effect为PreferNoSchedule【尽量不调度到该节点】;但只有该节点满足调度条件,因此都调度到了k8s-node01节点。

 

Pod单个容忍时(Tolerations)

记得把已有的污点清除,以免影响测验。

节点上的污点设置(Taints)

实现如下污点

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:check-nginx=web:PreferNoSchedule
3 k8s-node02 污点为:check-nginx=web:NoSchedule

 

污点添加操作如下:

1 kubectl taint nodes k8s-node01 check-nginx=web:PreferNoSchedule
2 kubectl taint nodes k8s-node02 check-nginx=web:NoSchedule

 

污点查看操作如下:

1 kubectl describe node k8s-master | grep 'Taints' -A 5
2 kubectl describe node k8s-node01 | grep 'Taints' -A 5
3 kubectl describe node k8s-node02 | grep 'Taints' -A 5

 

单个容忍示例

yaml文件

 1 [root@k8s-master taint]# pwd
 2 /root/k8s_practice/scheduler/taint
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# cat one_tolerations.yaml 
 5 apiVersion: apps/v1
 6 kind: Deployment
 7 metadata:
 8   name: one-tolerations-deploy
 9   labels:
10     app: onetolerations-deploy
11 spec:
12   replicas: 6
13   selector:
14     matchLabels:
15       app: myapp
16   template:
17     metadata:
18       labels:
19         app: myapp
20     spec:
21       containers:
22       - name: myapp-pod
23         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1
24         imagePullPolicy: IfNotPresent
25         ports:
26           - containerPort: 80
27       tolerations:
28       - key: "check-nginx"
29         operator: "Equal"
30         value: "web"
31         effect: "NoSchedule"

 

运行yaml文件

 1 [root@k8s-master taint]# kubectl apply -f one_tolerations.yaml 
 2 deployment.apps/one-tolerations-deploy created
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl get deploy -o wide
 5 NAME                     READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR
 6 one-tolerations-deploy   6/6     6            6           3s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp
 7 [root@k8s-master taint]# 
 8 [root@k8s-master taint]# kubectl get pod -o wide
 9 NAME                                      READY   STATUS    RESTARTS   AGE   IP            NODE         NOMINATED NODE   READINESS GATES
10 one-tolerations-deploy-5757d6b559-gbj49   1/1     Running   0          7s    10.244.2.73   k8s-node02   <none>           <none>
11 one-tolerations-deploy-5757d6b559-j9p6r   1/1     Running   0          7s    10.244.2.71   k8s-node02   <none>           <none>
12 one-tolerations-deploy-5757d6b559-kpk9q   1/1     Running   0          7s    10.244.2.72   k8s-node02   <none>           <none>
13 one-tolerations-deploy-5757d6b559-lsppn   1/1     Running   0          7s    10.244.4.65   k8s-node01   <none>           <none>
14 one-tolerations-deploy-5757d6b559-rx72g   1/1     Running   0          7s    10.244.4.66   k8s-node01   <none>           <none>
15 one-tolerations-deploy-5757d6b559-s8qr9   1/1     Running   0          7s    10.244.2.74   k8s-node02   <none>           <none>

由上可见,此时pod会尽量【优先】调度到k8s-node02节点,尽量不调度到k8s-node01节点。如果我们只有一个pod,那么会一直调度到k8s-node02节点。

 

Pod多个容忍时(Tolerations)

记得把已有的污点清除,以免影响测验。

节点上的污点设置(Taints)

实现如下污点

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:check-nginx=web:PreferNoSchedule, check-redis=memdb:NoSchedule
3 k8s-node02 污点为:check-nginx=web:NoSchedule, check-redis=database:NoSchedule

 

污点添加操作如下:

1 kubectl taint nodes k8s-node01 check-nginx=web:PreferNoSchedule
2 kubectl taint nodes k8s-node01 check-redis=memdb:NoSchedule
3 kubectl taint nodes k8s-node02 check-nginx=web:NoSchedule
4 kubectl taint nodes k8s-node02 check-redis=database:NoSchedule

 

污点查看操作如下:

1 kubectl describe node k8s-master | grep 'Taints' -A 5
2 kubectl describe node k8s-node01 | grep 'Taints' -A 5
3 kubectl describe node k8s-node02 | grep 'Taints' -A 5

 

多个容忍示例

yaml文件

 1 [root@k8s-master taint]# pwd
 2 /root/k8s_practice/scheduler/taint
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# cat multi_tolerations.yaml 
 5 apiVersion: apps/v1
 6 kind: Deployment
 7 metadata:
 8   name: multi-tolerations-deploy
 9   labels:
10     app: multitolerations-deploy
11 spec:
12   replicas: 6
13   selector:
14     matchLabels:
15       app: myapp
16   template:
17     metadata:
18       labels:
19         app: myapp
20     spec:
21       containers:
22       - name: myapp-pod
23         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1
24         imagePullPolicy: IfNotPresent
25         ports:
26           - containerPort: 80
27       tolerations:
28       - key: "check-nginx"
29         operator: "Equal"
30         value: "web"
31         effect: "NoSchedule"
32       - key: "check-redis"
33         operator: "Exists"
34         effect: "NoSchedule"

 

运行yaml文件

 1 [root@k8s-master taint]# kubectl apply -f multi_tolerations.yaml 
 2 deployment.apps/multi-tolerations-deploy created
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl get deploy -o wide
 5 NAME                       READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR
 6 multi-tolerations-deploy   6/6     6            6           5s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp
 7 [root@k8s-master taint]# 
 8 [root@k8s-master taint]# kubectl get pod -o wide
 9 NAME                                        READY   STATUS    RESTARTS   AGE   IP             NODE         NOMINATED NODE   READINESS GATES
10 multi-tolerations-deploy-776ff4449c-2csnk   1/1     Running   0          10s   10.244.2.171   k8s-node02   <none>           <none>
11 multi-tolerations-deploy-776ff4449c-4d9fh   1/1     Running   0          10s   10.244.4.116   k8s-node01   <none>           <none>
12 multi-tolerations-deploy-776ff4449c-c8fz5   1/1     Running   0          10s   10.244.2.173   k8s-node02   <none>           <none>
13 multi-tolerations-deploy-776ff4449c-nj29f   1/1     Running   0          10s   10.244.4.115   k8s-node01   <none>           <none>
14 multi-tolerations-deploy-776ff4449c-r7gsm   1/1     Running   0          10s   10.244.2.172   k8s-node02   <none>           <none>
15 multi-tolerations-deploy-776ff4449c-s8t2n   1/1     Running   0          10s   10.244.2.174   k8s-node02   <none>           <none>

由上可见,示例中的pod容忍为:check-nginx=web:NoSchedule;check-redis=:NoSchedule。因此pod会尽量调度到k8s-node02节点,尽量不调度到k8s-node01节点。

 

Pod容忍指定污点key的所有effects情况

记得把已有的污点清除,以免影响测验。

节点上的污点设置(Taints)

实现如下污点

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:check-redis=memdb:NoSchedule
3 k8s-node02 污点为:check-redis=database:NoSchedule

 

污点添加操作如下:

1 kubectl taint nodes k8s-node01 check-redis=memdb:NoSchedule
2 kubectl taint nodes k8s-node02 check-redis=database:NoSchedule

 

污点查看操作如下:

1 kubectl describe node k8s-master | grep 'Taints' -A 5
2 kubectl describe node k8s-node01 | grep 'Taints' -A 5
3 kubectl describe node k8s-node02 | grep 'Taints' -A 5

 

多个容忍示例

yaml文件

 1 [root@k8s-master taint]# pwd
 2 /root/k8s_practice/scheduler/taint
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# cat key_tolerations.yaml 
 5 apiVersion: apps/v1
 6 kind: Deployment
 7 metadata:
 8   name: key-tolerations-deploy
 9   labels:
10     app: keytolerations-deploy
11 spec:
12   replicas: 6
13   selector:
14     matchLabels:
15       app: myapp
16   template:
17     metadata:
18       labels:
19         app: myapp
20     spec:
21       containers:
22       - name: myapp-pod
23         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1
24         imagePullPolicy: IfNotPresent
25         ports:
26           - containerPort: 80
27       tolerations:
28       - key: "check-redis"
29         operator: "Exists"

 

运行yaml文件

 1 [root@k8s-master taint]# kubectl apply -f key_tolerations.yaml 
 2 deployment.apps/key-tolerations-deploy created
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl get deploy -o wide
 5 NAME                     READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR
 6 key-tolerations-deploy   6/6     6            6           21s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp
 7 [root@k8s-master taint]# 
 8 [root@k8s-master taint]# kubectl get pod -o wide
 9 NAME                                     READY   STATUS    RESTARTS   AGE   IP             NODE         NOMINATED NODE   READINESS GATES
10 key-tolerations-deploy-db5c4c4db-2zqr8   1/1     Running   0          26s   10.244.2.170   k8s-node02   <none>           <none>
11 key-tolerations-deploy-db5c4c4db-5qb5p   1/1     Running   0          26s   10.244.4.113   k8s-node01   <none>           <none>
12 key-tolerations-deploy-db5c4c4db-7xmt6   1/1     Running   0          26s   10.244.2.169   k8s-node02   <none>           <none>
13 key-tolerations-deploy-db5c4c4db-84rkj   1/1     Running   0          26s   10.244.4.114   k8s-node01   <none>           <none>
14 key-tolerations-deploy-db5c4c4db-gszxg   1/1     Running   0          26s   10.244.2.168   k8s-node02   <none>           <none>
15 key-tolerations-deploy-db5c4c4db-vlgh8   1/1     Running   0          26s   10.244.4.112   k8s-node01   <none>           <none>

由上可见,示例中的pod容忍为:check-nginx=:;仅需匹配node污点的key即可,污点的value和effect不需要关心。因此可以匹配k8s-node01、k8s-node02节点。

 

Pod容忍所有污点

记得把已有的污点清除,以免影响测验。

节点上的污点设置(Taints)

实现如下污点

1 k8s-master 污点为:node-role.kubernetes.io/master:NoSchedule 【k8s自带污点,直接使用,不必另外操作添加】
2 k8s-node01 污点为:check-nginx=web:PreferNoSchedule, check-redis=memdb:NoSchedule
3 k8s-node02 污点为:check-nginx=web:NoSchedule, check-redis=database:NoSchedule

 

污点添加操作如下:

1 kubectl taint nodes k8s-node01 check-nginx=web:PreferNoSchedule
2 kubectl taint nodes k8s-node01 check-redis=memdb:NoSchedule
3 kubectl taint nodes k8s-node02 check-nginx=web:NoSchedule
4 kubectl taint nodes k8s-node02 check-redis=database:NoSchedule

 

污点查看操作如下:

1 kubectl describe node k8s-master | grep 'Taints' -A 5
2 kubectl describe node k8s-node01 | grep 'Taints' -A 5
3 kubectl describe node k8s-node02 | grep 'Taints' -A 5

 

所有容忍示例

yaml文件

 1 [root@k8s-master taint]# pwd
 2 /root/k8s_practice/scheduler/taint
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# cat all_tolerations.yaml 
 5 apiVersion: apps/v1
 6 kind: Deployment
 7 metadata:
 8   name: all-tolerations-deploy
 9   labels:
10     app: alltolerations-deploy
11 spec:
12   replicas: 6
13   selector:
14     matchLabels:
15       app: myapp
16   template:
17     metadata:
18       labels:
19         app: myapp
20     spec:
21       containers:
22       - name: myapp-pod
23         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1
24         imagePullPolicy: IfNotPresent
25         ports:
26           - containerPort: 80
27       tolerations:
28       - operator: "Exists"

 

运行yaml文件

 1 [root@k8s-master taint]# kubectl apply -f all_tolerations.yaml 
 2 deployment.apps/all-tolerations-deploy created
 3 [root@k8s-master taint]# 
 4 [root@k8s-master taint]# kubectl get deploy -o wide
 5 NAME                     READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR
 6 all-tolerations-deploy   6/6     6            6           8s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp
 7 [root@k8s-master taint]# 
 8 [root@k8s-master taint]# kubectl get pod -o wide
 9 NAME                                      READY   STATUS    RESTARTS   AGE   IP             NODE         NOMINATED NODE   READINESS GATES
10 all-tolerations-deploy-566cdccbcd-4klc2   1/1     Running   0          12s   10.244.0.116   k8s-master   <none>           <none>
11 all-tolerations-deploy-566cdccbcd-59vvc   1/1     Running   0          12s   10.244.0.115   k8s-master   <none>           <none>
12 all-tolerations-deploy-566cdccbcd-cvw4s   1/1     Running   0          12s   10.244.2.175   k8s-node02   <none>           <none>
13 all-tolerations-deploy-566cdccbcd-k8fzl   1/1     Running   0          12s   10.244.2.176   k8s-node02   <none>           <none>
14 all-tolerations-deploy-566cdccbcd-s2pw7   1/1     Running   0          12s   10.244.4.118   k8s-node01   <none>           <none>
15 all-tolerations-deploy-566cdccbcd-xzngt   1/1     Running   0          13s   10.244.4.117   k8s-node01   <none>           <none>

后上可见,示例中的pod容忍所有的污点,因此pod可被调度到所有k8s节点。

 

相关阅读

1、官网:污点与容忍

2、Kubernetes K8S调度器kube-scheduler详解

3、Kubernetes K8S之affinity亲和性与反亲和性详解与示例

完毕!

 


 

———END———
如果觉得不错就关注下呗 (-^O^-) !