两类：

• 节点选择器：nodeSelector（给node打上标签，pod通过标签预选节点），nodeName
• 节点亲和调度：nodeAffinity

1、节点选择器(nodeSelector,nodeName)

[root@master ~]# kubectl explain pods.spec.nodeSelector[root@master schedule]# pwd/root/manifests/schedule[root@master schedule]# vim pod-demo.yaml apiVersion: v1kind: Podmetadata:  name: pod-demo  namespace: default  labels:    app: myapp    tier: frontend  annotations:     mageedu.com/created-by: "cluster admin"spec:  containers:  - name: myapp    image: ikubernetes/myapp:v1  nodeSelector:    #节点选择器    disktype: ssd    #该pod运行在有disktype=ssd标签的node节点上

[root@master schedule]# kubectl apply -f pod-demo.yaml pod/pod-demo created[root@master schedule]# kubectl get pods -o wideNAME       READY   STATUS    RESTARTS   AGE     IP           NODE     NOMINATED NODE   READINESS GATESpod-demo   1/1     Running   0          8m13s   10.244.1.6   node01   
                
      
       [root@master schedule]# kubectl get nodes --show-labels |grep node01node01   Ready    
       
           76d   v1.13.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/hostname=node01#可见新创建的pod已经运行在node01上了，因为node01上有disktype=ssd标签；
       
      
     

接下来我们给node02打上标签，修改一下资源定义清单文件，再创建pod：

将node02打上标签，pod资源清单里面的节点选择器里，改为和node02一样的标签；

[root@master schedule]# kubectl delete -f pod-demo.yaml[root@master ~]# kubectl label nodes node02 disktype=harddisknode/node02 labeled[root@master schedule]# vim pod-demo.yaml apiVersion: v1kind: Podmetadata:  name: pod-demo  namespace: default  labels:    app: myapp    tier: frontend  annotations:    mageedu.com/created-by: "cluster admin"spec:  containers:  - name: myapp    image: ikubernetes/myapp:v1  nodeSelector:    disktype: harddisk        [root@master schedule]# kubectl get nodes --show-labels |grep node02node02   Ready    
     
         76d   v1.13.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=harddisk,kubernetes.io/hostname=node02[root@master schedule]# kubectl get pods -o wideNAME       READY   STATUS    RESTARTS   AGE    IP           NODE     NOMINATED NODE   READINESS GATESpod-demo   1/1     Running   0          104s   10.244.2.5   node02   
                 
       
      
     

可见pod已经运行在node02上了；

2、节点亲和度调度

[root@master scheduler]# kubectl explain pods.spec.affinity[root@master scheduler]# kubectl explain pods.spec.affinity.nodeAffinitypreferredDuringSchedulingIgnoredDuringExecution:软亲和，requiredDuringSchedulingIgnoredDuringExecution:硬亲和，表示必须满足[root@master ~]# kubectl explain pods.spec.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms.matchExpressions#硬亲和性[root@master schedule]# vim pod-nodeaffinity-demo.yaml apiVersion: v1kind: Podmetadata:  name: pod-node-affinity-demo  namespace: default  labels:    app: myapp    tier: frontendspec:  containers:  - name: myapp    image: ikubernetes/myapp:v1  affinity:    nodeAffinity:      requiredDuringSchedulingIgnoredDuringExecution:        nodeSelectorTerms:        - matchExpressions:          - key: zone            operator: In            values:            - foo            - bar[root@master schedule]# kubectl apply -f pod-nodeaffinity-demo.yaml pod/pod-node-affinity-demo created[root@master schedule]# kubectl get pods    NAME                     READY   STATUS    RESTARTS   AGEpod-node-affinity-demo   0/1     Pending   0          76s #此时pod是Pending, 是因为没有节点满足条件；

下面我们再创建一个软亲和性的pod：

#软亲和性，就算没有符合条件的节点，也会找一个勉强运行；[root@master schedule]# kubectl delete -f pod-nodeaffinity-demo.yaml pod "pod-node-affinity-demo" deleted[root@master schedule]# vim pod-nodeaffinity-demo2.yaml apiVersion: v1kind: Podmetadata:  name: pod-node-affinity-demo2  namespace: default  labels:    app: myapp    tier: frontendspec:  containers:  - name: myapp    image: ikubernetes/myapp:v1  affinity:    nodeAffinity:      preferredDuringSchedulingIgnoredDuringExecution:      - preference:          matchExpressions:          - key: zone            operator: In            values:            - foo            - bar        weight: 60[root@master schedule]# kubectl apply -f pod-nodeaffinity-demo2.yaml pod/pod-node-affinity-demo2 created[root@master schedule]# kubectl get pods    #可见pod已经运行了NAME                      READY   STATUS    RESTARTS   AGEpod-node-affinity-demo2   1/1     Running   0          74spod-node-affinity-demo-2 运行起来了，因为这个pod我们是定义的软亲和性，即使没有符合条件的及诶单，也会找个节点让Pod运行起来

 

3、pod亲和性调度

比如在机房中，我们可以将一个机柜中的机器都打上标签，让pod调度的时候，对此机柜有亲和性；

或者将机柜中某几台机器打上标签，让pod调度的时候，对这几个机器有亲和性；

#查看资源定义清单字段[root@master ~]# kubectl explain pods.spec.affinity.podAffinityFIELDS:   preferredDuringSchedulingIgnoredDuringExecution    <[]Object>  #软亲和  requiredDuringSchedulingIgnoredDuringExecution    <[]Object>  #硬亲和[root@master ~]# kubectl explain pods.spec.affinity.podAffinity.requiredDuringSchedulingIgnoredDuringExecutionFIELDS:   labelSelector    

pod硬亲和性调度：

[root@master ~]# kubectl get nodes --show-labelsNAME     STATUS   ROLES    AGE   VERSION   LABELSmaster   Ready    master   77d   v1.13.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/hostname=master,node-role.kubernetes.io/master=node01   Ready    
     
         77d   v1.13.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/hostname=node01node02   Ready    
      
          76d   v1.13.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=harddisk,kubernetes.io/hostname=node02#资源定义清单[root@master schedule]# vim pod-requieed-affinity-demo.yaml apiVersion: v1kind: Podmetadata:  name: pod-first  namespace: default  labels:    app: myapp    tier: frontendspec:  containers:  - name: myapp    image: ikubernetes/myapp:v1---apiVersion: v1kind: Podmetadata:  name: pod-second  namespace: default  labels:    app: backend    tier: dbspec:  containers:  - name: busybox    #前面的-号表示这是一个列表格式的,也可以用中括号表示    image: busybox:latest    imagePullPolicy: IfNotPresent    command: ["sh","-c","sleep 3600"]  affinity:    podAffinity:      requiredDuringSchedulingIgnoredDuringExecution:    #硬亲和性      - labelSelector:          matchExpressions:          - {key: app,operator: In,values: ["myapp"]}   #意思是当前这个pod要跟一个有着标签app=myapp(要和上面pod-first的metadata里面的标签一致)的pod在一起        topologyKey: kubernetes.io/hostname    #匹配的节点key是kubernetes.io/hostname#创建[root@master schedule]# kubectl apply -f pod-requieed-affinity-demo.yaml pod/pod-first createdpod/pod-second created[root@master schedule]# kubectl get pods -o wideNAME         READY   STATUS    RESTARTS   AGE     IP            NODE     NOMINATED NODE   READINESS GATESpod-first    1/1     Running   0          3m25s   10.244.2.9    node02   
                  
        
         pod-second   1/1     Running   0          3m25s   10.244.2.10   node02   
                    
          
           #可以看到我们的两个pod都运行在同一个节点了，这是因为pod-second会和pod-first运行在同一个节点上，pod-second依赖于pod-first；
          
         
        
       
      
     

4、pod反亲和性调度

[root@master ~]# kubectl explain pods.spec.affinity.podAntiAffinity.requiredDuringSchedulingIgnoredDuringExecution.labelSelectorFIELDS:   matchExpressions    <[]Object>   matchLabels    
     
      [root@master schedule]# kubectl delete -f pod-requieed-affinity-demo.yaml    #删掉刚才的pod#资源定义清单[root@master schedule]# vim pod-requieed-Anti-affinity-demo.yaml apiVersion: v1kind: Podmetadata:  name: pod-first  namespace: default  labels:    app: myapp    tier: frontendspec:  containers:  - name: myapp    image: ikubernetes/myapp:v1---apiVersion: v1kind: Podmetadata:  name: pod-second  namespace: default  labels:    app: backend    tier: dbspec:  containers:  - name: busybox    image: busybox:latest    imagePullPolicy: IfNotPresent    command: ["sh","-c","sleep 3600"]  affinity:    podAntiAffinity:      requiredDuringSchedulingIgnoredDuringExecution:      - labelSelector:          matchExpressions:          - {key: app,operator: In,values: ["myapp"]}        topologyKey: kubernetes.io/hostname#创建[root@master schedule]# kubectl apply -f  pod-requieed-Anti-affinity-demo.yaml pod/pod-first createdpod/pod-second created[root@master schedule]# kubectl get pods -o wideNAME         READY   STATUS    RESTARTS   AGE   IP            NODE     NOMINATED NODE   READINESS GATESpod-first    1/1     Running   0          53s   10.244.1.7    node01   
                 
       
        pod-second   1/1     Running   0          53s   10.244.2.11   node02   
                   
         
          #可见pod-first和pod-second就不会被调度到同一个节点上；
         
        
       
      
     

下面可以给两个节点打相同的标签，因为pod调度策略是podAntiAffinity反亲和性，所以pod-first和pod-second不能同时运行在标有zone标签的节点上；

最终出现的情况就是有一个pod-first能成功运行，而另外一个pod-second因为是反亲和的，没有节点可以运行而处于pending状态；

#打标，相同的标签[root@master ~]# kubectl label nodes node01 zone=foonode/node01 labeled[root@master ~]# kubectl label nodes node02 zone=foo[root@master schedule]# kubectl delete -f  pod-requieed-Anti-affinity-demo.yaml    #删掉pod#资源定义定义清单[root@master schedule]# vim pod-requieed-Anti-affinity-demo.yaml apiVersion: v1kind: Podmetadata:  name: pod-first  namespace: default  labels:    app: myapp    tier: frontendspec:  containers:  - name: myapp    image: ikubernetes/myapp:v1---apiVersion: v1kind: Podmetadata:  name: pod-second  namespace: default  labels:    app: backend    tier: dbspec:  containers:  - name: busybox    image: busybox:latest    imagePullPolicy: IfNotPresent    command: ["sh","-c","sleep 3600"]  affinity:    podAntiAffinity:      requiredDuringSchedulingIgnoredDuringExecution:      - labelSelector:          matchExpressions:          - {key: app,operator: In,values: ["myapp"]}        topologyKey: zone            #节点标签改为zone#创建[root@master schedule]# kubectl apply -f  pod-requieed-Anti-affinity-demo.yaml pod/pod-first createdpod/pod-second created[root@master schedule]# kubectl get pods -o wideNAME         READY   STATUS    RESTARTS   AGE   IP            NODE     NOMINATED NODE   READINESS GATESpod-first    1/1     Running   0          4s    10.244.2.12   node02   
                
      
       pod-second   0/1     Pending   0          4s    
               
           
                    
          
           #可见pod-first能成功运行，而pod-second因为是反亲和的，没有节点可以运行而处于pending状态；
          
         
        
       
      
     

 

5、污点调度

污点调度是让节点来选择哪些pod能运行在其上面，污点(taints)用在节点上，容忍度(Tolerations )用在pod上；

污点定义：

[root@master ~]# kubectl explain nodes.spec.taints    #taints：定义节点的污点FIELDS:   effect    
     
       -required-   #表示当pod不能容忍节点上污点时的行为是什么，主要有以下三种行为：   {NoSchedule:仅影响调度过程，不影响现存pod。没调度过来的就调度不过来了。如果对节点新加了污点，那么对节点上现存的Pod没有影响。    NoExecute:既影响调度过程，也影响现存Pod，没调度过来的就调度不过来了，如果对节点新加了污点，那么对现存的pod对象将会被驱逐    PreferNoSchedule:不能容忍就不能调度过来，但是实在没办法也是能调度过来的。对节点新加了污点，那么对节点上现存的pod没有影响。}   key    
      
        -required-   timeAdded    
       
           value    
        
         #查看节点的污点[root@master ~]# kubectl describe node node01 |grep TaintsTaints:             
         
          [root@master ~]# kubectl describe node node02 |grep TaintsTaints:             
          
           #查看pod的容忍度[root@master ~]# kubectl describe pods kube-apiserver-master -n kube-system |grep TolerationsTolerations: :NoExecute[root@master ~]# kubectl taint -h | grep -A 1 Usage #给节点打污点的方式Usage: kubectl taint NODE NAME KEY_1=VAL_1:TAINT_EFFECT_1 ... KEY_N=VAL_N:TAINT_EFFECT_N [options]污点和容忍度都是自定义的键值对形式；下面给node1打上污点node-type=production:NoSchedule：[root@master ~]# kubectl taint node node01 node-type=production:NoSchedulenode/node01 tainted#pod资源定义清单，此文件没有定义容忍度，但是node01有污点，pod应该都会运行在node02上；[root@master schedule]# vim deploy-demo.yaml apiVersion: apps/v1kind: Deploymentmetadata: name: myapp-deploy namespace: defaultspec: replicas: 3 selector: matchLabels: app: myapp release: canary template: metadata: labels: app: myapp release: canary spec: containers: - name: myapp image: ikubernetes/myapp:v2 ports: - name: http containerPort: 80#创建[root@master schedule]# kubectl apply -f deploy-demo.yaml deployment.apps/myapp-deploy created#可见pod都运行在了node02上[root@master schedule]# kubectl get pods -o wideNAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATESmyapp-deploy-6b56d98b6b-52hth 1/1 Running 0 9s 10.244.2.15 node02 
            
            
             myapp-deploy-6b56d98b6b-dr224 1/1 Running 0 9s 10.244.2.14 node02 
              
              
               myapp-deploy-6b56d98b6b-z278x 1/1 Running 0 9s 10.244.2.13 node02 
                
                
               
              
             
            
           
          
         
        
       
      
     

容忍度定义：

[root@master ~]# kubectl explain pods.spec.tolerationsFIELDS:   effect    
     
         key    
      
          operator    
       
            #两个值：Exists表示只要节点有这个污点的key，pod都能容忍，值是什么都行；Equal表示只要节点必须精确匹配污点的key和value才能容忍；   tolerationSeconds    
        
               #表示宽限多长时间pod才会被驱逐   value    
         
          [root@master ~]# kubectl taint node node02 node-type=dev:NoExecute    #给node02打上另一个标签node/node02 tainted[root@master schedule]# kubectl delete -f deploy-demo.yaml#资源定义清单[root@master schedule]# vim deploy-demo.yaml apiVersion: apps/v1kind: Deploymentmetadata:  name: myapp-deploy  namespace: defaultspec:  replicas: 3  selector:    matchLabels:      app: myapp      release: canary  template:    metadata:      labels:        app: myapp        release: canary    spec:      containers:      - name: myapp        image: ikubernetes/myapp:v2        ports:        - name: http          containerPort: 80      tolerations:      - key: "node-type"        operator: "Equal"    #要精确匹配污点键值        value: "production"        effect: "NoSchedule"#创建pod[root@master schedule]# kubectl apply -f deploy-demo.yaml deployment.apps/myapp-deploy created[root@master schedule]# kubectl get pods -o wideNAME                            READY   STATUS    RESTARTS   AGE   IP            NODE     NOMINATED NODE   READINESS GATESmyapp-deploy-779c578779-5vkbw   1/1     Running   0          12s   10.244.1.12   node01   
           
           
            myapp-deploy-779c578779-bh9td 1/1 Running 0 12s 10.244.1.11 node01 
             
             
              myapp-deploy-779c578779-dn52p 1/1 Running 0 12s 10.244.1.13 node01 
               
               
                #可见pod都运行在了node01上，因为我们设置了pod能容忍node01的污点；
               
              
             
            
           
          
         
        
       
      
     

下面我们把operator: "Equal"改成operator: "Exists"

Exists表示只要节点有这个污点的key，pod都能容忍，值是什么都行；

[root@master schedule]# kubectl delete -f deploy-demo.yaml[root@master schedule]# vim deploy-demo.yaml apiVersion: apps/v1kind: Deploymentmetadata:  name: myapp-deploy  namespace: defaultspec:  replicas: 3  selector:    matchLabels:      app: myapp      release: canary  template:    metadata:      labels:        app: myapp        release: canary    spec:      containers:      - name: myapp        image: ikubernetes/myapp:v2        ports:        - name: http          containerPort: 80      tolerations:      - key: "node-type"        operator: "Exists"          value: ""        effect: ""       #不设置行为#创建[root@master schedule]# kubectl apply -f deploy-demo.yaml deployment.apps/myapp-deploy create[root@master schedule]# kubectl get pods -o wideNAME                            READY   STATUS    RESTARTS   AGE   IP            NODE     NOMINATED NODE   READINESS GATESmyapp-deploy-69b95476c8-bfpgj   1/1     Running   0          13s   10.244.2.20   node02   
                
      
       myapp-deploy-69b95476c8-fhwbd   1/1     Running   0          13s   10.244.1.17   node01   
                  
        
         myapp-deploy-69b95476c8-tzzlx   1/1     Running   0          13s   10.244.2.19   node02   
                    
          
           #可见，node01 node02上面都有pod了； effect：不设置表示什么行为都能容忍；
          
         
        
       
      
     

最后可以去除节点上的污点：

#去除污点命令，删除指定key上所有的effect[root@master ~]# kubectl taint node node02 node-type-node/node02 untainted[root@master ~]# kubectl taint node node01 node-type-node/node01 untainted