k8s调度器

节点硬亲和度：Pod调度时必须满足规划，如不满足Pod状态将是Penging状态。

十载的公安网站建设经验，针对设计、前端、开发、售后、文案、推广等六对一服务，响应快，48小时及时工作处理。网络营销推广的优势是能够根据用户设备显示端的尺寸不同，自动调整公安建站的显示方式，使网站能够适用不同显示终端，在浏览器中调整网站的宽度，无论在任何一种浏览器上浏览网站，都能展现优雅布局与设计，从而大程度地提升浏览体验。创新互联建站从事“公安网站设计”,“公安网站推广”以来，每个客户项目都认真落实执行。

节点软亲和度：Pod调度时先按规则进行调度，如不满足规则也会找一个不匹配的节点运行。

Pod硬亲和度，软亲和度与节点硬亲和度，软亲和度相似。

1.节点硬亲和度(节点不满足规则)

[root@k8s01 yaml]# cat pod-affinity01.yaml

apiVersion: v1
kind: Pod
metadata:
  name: pod-01
spec:
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - {key: zone,operator: In,values: ["one"]}   --节点zone(键)标签必须是one(值)
  containers:
  - name: pod-01
    image: nginx:latest
    imagePullPolicy: Never

[root@k8s01 yaml]# kubectl apply -f pod-affinity01.yaml

pod/pod-01 created

[root@k8s01 yaml]# kubectl get pods --show-labels

NAME    READY    STATUS      RESTARTS      AGE      LABELS

pod-01       0/1        Pending                 0            103s         --node节点不满足pod标签

[root@k8s01 yaml]# kubectl describe pods pod-01

。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。

Events:

Type Reason Age From Message

---- ------ ---- ---- -------

Warning FailedScheduling default-scheduler 0/3 nodes are available: 3 node(s) didn't match node selector.

Warning FailedScheduling default-scheduler 0/3 nodes are available: 3 node(s) didn't match node selector.

[root@k8s01 yaml]#

- matchExpressions:
  - {key: zone,operator: In,values: ["one"]} --zone是键，operator：是表达式，one是值
  - {key: zone,operator: In,values: ["one","two"]} --In 是包含
  - {key: ssd,operator: Exists,values: []} --Exists是存在

2.节点硬亲和度(节点满足规则)

[root@k8s01 yaml]# kubectl label node k8s02 zone=one     --在k8s02节点创建标签

node/k8s02 labeled

[root@k8s01 yaml]# kubectl get pods --show-labels -o wide

NAME    READY    STATUS    RESTARTS    AGE    IP    NODE    NOMINATED    NODE    READINESS   GATES LABELS

pod-01      1/1            Running       0         6m13s          10.244.1.37         k8s02              

[root@k8s01 yaml]#

3.节点软亲和度

[root@k8s01 yaml]# cat pod-affinity02.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: pod-02
spec:
  replicas: 3
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      name: myapp-pod
      labels:
        app: myapp
    spec:
      affinity:
        nodeAffinity:
          preferredDuringSchedulingIgnoredDuringExecution:
          - weight: 60    --60%的Pod会调度到zone标签，one值这个节点上
            preference:
              matchExpressions:
              - {key: zone,operator: In,values: ["one"]}
      containers:
      - name: pod-02
        image: nginx:latest
        imagePullPolicy: Never

[root@k8s01 yaml]# kubectl apply -f pod-affinity02.yaml

deployment.apps/pod-02 created

[root@k8s01 yaml]# kubectl get pods --show-labels -o wide    --满足标签可创建，不满足也可创建

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS

pod-02-77d87986b-9bzg5 1/1 Running 0 16s 10.244.1.39 k8s02 app=myapp,pod-template-hash=77d87986b

pod-02-77d87986b-dckjq 1/1 Running 0 16s 10.244.2.42 k8s03 app=myapp,pod-template-hash=77d87986b

pod-02-77d87986b-z7v47 1/1 Running 0 16s 10.244.1.38 k8s02 app=myapp,pod-template-hash=77d87986b

[root@k8s01 yaml]#

4. Pod硬亲和度(节点不满足规则)

[root@k8s01 yaml]# kubectl  get pods --show-labels
NAME                     READY   STATUS    RESTARTS   AGE   LABELS
nginx                    1/1     Running   0          4s    app=web
[root@k8s01 yaml]# cat pod-affinity03.yaml 

apiVersion: v1
kind: Pod
metadata:
  name: pod-01
spec:
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - {key: app,operator: In,values: ["web1"]}   --标签值不满足
        topologyKey: kubernetes.io/hostname
  containers:
  - name: pod-01
    image: nginx:latest
    imagePullPolicy: Never

[root@k8s01 yaml]# kubectl apply -f  pod-affinity03.yaml
pod/pod-01 created
[root@k8s01 yaml]# kubectl  get pods --show-labels -o wide
NAME     READY   STATUS    RESTARTS   AGE    IP      NODE     NOMINATED NODE   READINESS GATES   LABELS
nginx                    1/1     Running   0          8m9s   10.244.1.42   k8s02                           app=web
pod-01                  0/1     Pending   0          28s                                     
[root@k8s01 yaml]#

5.Pod硬亲和度 (节点满足规则)

[root@k8s01 yaml]# kubectl  get pods --show-labels      --查看pod标签
NAME                     READY   STATUS    RESTARTS   AGE   LABELS
nginx                    1/1     Running   0          4s    app=web
[root@k8s01 yaml]# cat pod-affinity04.yaml 

apiVersion: v1
kind: Pod
metadata:
  name: pod-01
spec:
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - {key: app,operator: In,values: ["web"]}
        topologyKey: kubernetes.io/hostname
  containers:
  - name: pod-01
    image: nginx:latest
    imagePullPolicy: Never

[root@k8s01 yaml]# kubectl apply -f  pod-affinity04.yaml
pod/pod-01 created
[root@k8s01 yaml]# kubectl  get pods --show-labels -o wide       --创建的新Pod会找app=web的标签
NAME     READY   STATUS    RESTARTS   AGE     IP     NODE    NOMINATED NODE   READINESS GATES   LABELS
nginx                    1/1     Running   0          4m14s   10.244.1.42   k8s02                          app=web
pod-01                  1/1     Running   0          17s     10.244.1.43   k8s02                         
[root@k8s01 yaml]#

6.Pod软亲和度

[root@k8s01 yaml]# kubectl  get pods --show-labels -o wide
NAME      READY   STATUS    RESTARTS   AGE     IP      NODE     NOMINATED NODE   READINESS GATES   LABELS
nginx                    1/1     Running   0          14m     10.244.1.42   k8s02                           app=web

[root@k8s01 yaml]# cat pod-affinity04.yaml 

apiVersion: apps/v1
kind: Deployment
metadata:
  name: pod-02
spec:
  replicas: 3
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      name: myapp-pod
      labels:
        app: myapp
    spec:
      affinity:
        podAffinity:
          preferredDuringSchedulingIgnoredDuringExecution:
          - weight: 60
            podAffinityTerm:
              labelSelector:
                matchExpressions:
                - {key: app,operator: In,values: ["web12"]}
              topologyKey: zone
      containers:
      - name: pod-02
        image: nginx:latest
        imagePullPolicy: Never

[root@k8s01 yaml]# kubectl  get pods --show-labels -o wide     --不满足Pod软亲和度也可创建
NAME      READY   STATUS    RESTARTS   AGE     IP    NODE     NOMINATED NODE   READINESS GATES   LABELS
nginx                    1/1     Running   0          16m     10.244.1.42   k8s02               app=web
pod-02-6f96fbfdf-52gbf   1/1     Running   0   11s     10.244.2.43   k8s03           app=myapp,pod-template-hash=6f96fbfdf
pod-02-6f96fbfdf-dl5z5   1/1     Running   0    11s     10.244.1.44   k8s02          app=myapp,pod-template-hash=6f96fbfdf
pod-02-6f96fbfdf-f8bzn   1/1     Running   0    11s     10.244.0.55   k8s01          app=myapp,pod-template-hash=6f96fbfdf
[root@k8s01 yaml]#

7.定义污点和容忍度

NoSchedule：不能容忍此污点的新Pod对像，不能调度到当前节点，强制性。

PreferNoSchedule：新Pod尽量不调度到此节点，若无其它节点可以调度，也允许使用当前节点调度，柔和性。

NoExecute：新Pod不能调度到当前节点，如果当前节点存在Pod也会被驱逐，强制性。

[root@k8s01 yaml]# kubectl  describe node k8s02 | grep -i taints     --查看污点
Taints:            
[root@k8s01 yaml]# kubectl taint node k8s02 node-type=production:NoSchedule    --为k8s02节点打污点

node/k8s02 tainted
[root@k8s01 yaml]# kubectl  describe node k8s02 | grep -i taints
Taints:             node-type=production:NoSchedule
[root@k8s01 yaml]# cat pod-affinity05.yaml 

apiVersion: apps/v1
kind: Deployment
metadata:
  name: pod-02
spec:
  replicas: 3
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      name: myapp-pod
      labels:
        app: myapp
    spec:
      containers:
      - name: pod-02
        image: nginx:latest
        imagePullPolicy: Never

[root@k8s01 yaml]# kubectl apply -f  pod-affinity05.yaml
deployment.apps/pod-02 created
[root@k8s01 yaml]# kubectl  get pods -o wide --show-labels     --k8s02节点已打污点，所以不能调度Pod对像
NAME     READY   STATUS    RESTARTS   AGE   IP       NODE    NOMINATED NODE   READINESS GATES   LABELS
pod-02-5c54dc6489-j7tsh   1/1     Running   0          29s   10.244.0.56   k8s01                          app=myapp,pod-template-hash=5c54dc6489
pod-02-5c54dc6489-sk5dm   1/1     Running   0          29s   10.244.2.44   k8s03                          app=myapp,pod-template-hash=5c54dc6489
pod-02-5c54dc6489-sn5wd   1/1     Running   0          29s   10.244.2.45   k8s03                          app=myapp,pod-template-hash=5c54dc6489

[root@k8s01 yaml]

8.删除污点

[root@k8s01 yaml]# kubectl  taint node k8s02 node-type:NoSchedule-      --删除 NoSchedule污点

node/k8s02 untainted

[root@k8s01 yaml]# kubectl taint node k8s02 node-type=production:PreferNoSchedule      --打上污点
node/k8s02 tainted
[root@k8s01 yaml]# kubectl  describe node k8s02 | grep -i node-type
Taints:             node-type=production:PreferNoSchedule
[root@k8s01 yaml]# kubectl taint node k8s02 node-type:PreferNoSchedule-    --删除 PreferNoSchedule污点

node/k8s02 untainted
[root@k8s01 yaml]# kubectl  describe node k8s02 | grep -i node-type

[root@k8s01 yaml]# kubectl taint node k8s02 node-type-                  --删除所有污点
node/k8s02 untainted
[root@k8s01 yaml]# kubectl  describe node k8s02 | grep -i node-type
[root@k8s01 yaml]#

9. 驱逐节点上的Pod

[root@k8s01 yaml]# kubectl cordon k8s03    --将新创建的pod不能调度到k8s03节点，以前的Pod不受影响。
node/k8s03 cordoned
[root@k8s01 yaml]# kubectl  get nodes
NAME    STATUS                     ROLES    AGE   VERSION
k8s01   Ready                      master   71d   v1.16.0
k8s02   Ready                         70d   v1.16.0
k8s03   Ready, SchedulingDisabled      30d   v1.16.0

[root@k8s01 yaml]# kubectl  get pods -o wide
NAME       READY   STATUS    RESTARTS   AGE   IP      NODE    NOMINATED NODE   READINESS GATES
pod-02-5c54dc6489-442kx   1/1     Running   0          12m   10.244.2.48   k8s03             
pod-02-5c54dc6489-92l8m   1/1     Running   0          12m   10.244.2.49   k8s03             
pod-02-5c54dc6489-k4bc7   1/1     Running   0          12m   10.244.0.58   k8s01             
[root@k8s01 yaml]# kubectl uncordon k8s03     --释放调度规则
node/k8s03 uncordoned
[root@k8s01 yaml]# kubectl  get nodes
NAME    STATUS   ROLES    AGE   VERSION
k8s01   Ready    master   71d   v1.16.0
k8s02   Ready       70d   v1.16.0
k8s03   Ready       30d   v1.16.0
[root@k8s01 yaml]# kubectl drain k8s03 --ignore-daemonsets     --去除k8s03节点上的Pod
node/k8s03 already cordoned
WARNING: ignoring DaemonSet-managed Pods: kube-system/kube-flannel-ds-amd64-cg795, kube-system/kube-proxy-h5dkf
evicting pod "pod-02-5c54dc6489-92l8m"
evicting pod "pod-02-5c54dc6489-442kx"
pod/pod-02-5c54dc6489-92l8m evicted
pod/pod-02-5c54dc6489-442kx evicted
node/k8s03 evicted
[root@k8s01 yaml]# kubectl  get nodes
NAME    STATUS               ROLES    AGE   VERSION
k8s01   Ready                      master   71d   v1.16.0
k8s02   Ready                         70d   v1.16.0
k8s03   Ready, SchedulingDisabled      30d   v1.16.0
[root@k8s01 yaml]# kubectl  get pods -o wide         --以前k8s03节点的Pod全部转移到其它节点上
NAME           READY   STATUS    RESTARTS   AGE   IP            NODE    NOMINATED NODE   READINESS GATES
pod-02-5c54dc6489-k4bc7   1/1     Running   0          14m   10.244.0.58   k8s01             
pod-02-5c54dc6489-mxk46   1/1     Running   0          25s   10.244.1.46   k8s02             
pod-02-5c54dc6489-vmb8l   1/1     Running   0          25s   10.244.1.45   k8s02             
[root@k8s01 yaml]# kubectl uncordon k8s03     --恢复调度
node/k8s03 uncordoned
[root@k8s01 yaml]# kubectl  get nodes
NAME    STATUS   ROLES    AGE   VERSION
k8s01   Ready    master   71d   v1.16.0
k8s02   Ready       70d   v1.16.0
k8s03   Ready       30d   v1.16.0
[root@k8s01 yaml]#