企业运维----Docker-kubernetes-调度

最新推荐文章于 2024-05-04 06:00:00 发布

oam_chen

最新推荐文章于 2024-05-04 06:00:00 发布

阅读量152

点赞数

分类专栏： Docker 笔记 k8s 文章标签：运维 docker k8s

本文链接：https://blog.csdn.net/weixin_56892849/article/details/119416684

版权

笔记同时被 3 个专栏收录

27 篇文章 0 订阅

订阅专栏

Docker

21 篇文章 0 订阅

订阅专栏

k8s

16 篇文章 0 订阅

订阅专栏

k8s调度概述

• 调度器通过 kubernetes 的 watch 机制来发现集群中新创建且尚未被调度到
Node 上的 Pod。调度器会将发现的每一个未调度的 Pod 调度到一个合适的 Node
上来运行。
• kube-scheduler 是 Kubernetes 集群的默认调度器,并且是集群控制面的一部分。
如果你真的希望或者有这方面的需求,kube-scheduler 在设计上是允许你自己写
一个调度组件并替换原有的 kube-scheduler。
• 在做调度决定时需要考虑的因素包括:单独和整体的资源请求、硬件/软件/策略限
制、亲和以及反亲和要求、数据局域性、负载间的干扰等等。
• 默认策略可以参考:
https://kubernetes.io/zh/docs/concepts/scheduling/kube-scheduler/
• 调度框架:
https://kubernetes.io/zh/docs/concepts/configuration/scheduling-
framework/

nodename节点选择约束

• nodeName 是节点选择约束的最简单方法,但一般不推荐。如果 nodeName 在
PodSpec 中指定了,则它优先于其他的节点选择方法。
• 使用 nodeName 来选择节点的一些限制:
• 如果指定的节点不存在。
• 如果指定的节点没有资源来容纳 pod,则pod 调度失败。
• 云环境中的节点名称并非总是可预测或稳定的。

[root@server2 ~]# mkdir schedu
[root@server2 ~]# cd schedu/
[root@server2 schedu]# vim pod1.yml
[root@server2 schedu]# cat pod1.yml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
spec:
  containers:

  - name: nginx
    image: nginx
  nodeName: server2

[root@server2 schedu]# kubectl apply -f pod1.yml 
pod/nginx created
[root@server2 schedu]# kubectl get pod
NAME                       READY   STATUS    RESTARTS   AGE
nginx                      1/1     Running   0          13s
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE     IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      1/1     Running   0          5m38s   10.244.0.20    server2   <none>           <none>

nodeSelector节点选择约束

拉起容器查看处于pending状态是因为未识别到指定标签

[root@server2 schedu]# kubectl delete -f pod1.yml 
pod "nginx" deleted
[root@server2 schedu]# vim pod2.yaml 
[root@server2 schedu]# cat pod2.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:

  - name: nginx
    image: nginx
    imagePullPolicy: IfNotPresent 
  nodeSelector:
    disktype: ssd

[root@server2 schedu]# kubectl apply -f pod2.yaml 
pod/nginx created
[root@server2 schedu]# kubectl get pod
NAME                       READY   STATUS    RESTARTS   AGE
nginx                      0/1     Pending   0          8s

给server3添加标签容器running

[root@server2 schedu]# kubectl label nodes server3 disktype=ssd
node/server2 labeled
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      1/1     Running   0          7m    10.244.5.103   server3   <none>           <none>

查看server3标签 disktype=ssd

[root@server2 schedu]# kubectl get node --show-labels 
NAME      STATUS   ROLES                  AGE   VERSION   LABELS
server2   Ready    control-plane,master   11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=server2,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=,node.kubernetes.io/exclude-from-external-load-balancers=
server3   Ready    <none>                 11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=server3,kubernetes.io/os=linux
server4   Ready    <none>                 11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=server4,kubernetes.io/os=linux

亲和与反亲和

• nodeSelector 提供了一种非常简单的方法来将 pod 约束到具有特定标签的
节点上。亲和/反亲和功能极大地扩展了你可以表达约束的类型。
• 你可以发现规则是“软”/“偏好”,而不是硬性要求,因此,如果调度器无
法满足该要求,仍然调度该 pod
• 你可以使用节点上的 pod 的标签来约束,而不是使用节点本身的标签,来允
许哪些 pod 可以或者不可以被放置在一起。

节点亲和

• requiredDuringSchedulingIgnoredDuringExecution 必须满足
• preferredDuringSchedulingIgnoredDuringExecution 倾向满足
• IgnoreDuringExecution 表示如果在Pod运行期间Node的标签发生变化,
导致亲和性策略不能满足,则继续运行当前的Pod。
• 参考:https://kubernetes.io/zh/docs/concepts/configuration/assign-
pod-node/

节点反亲和

指定pod node-affinity不在server4上部署

[root@server2 schedu]# kubectl delete -f pod2.yaml 
pod "nginx" deleted
[root@server2 schedu]# vim pod3.yaml 
[root@server2 schedu]# cat pod3.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: node-affinity
spec:
  containers:
  - name: nginx
    image: nginx
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
           nodeSelectorTerms:
           - matchExpressions:
             - key: kubernetes.io/hostname
               operator: NotIn
               values:
               - server4
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 1
        preference:
          matchExpressions:
          - key: disktype
            operator: In
            values:
            - ssd 

[root@server2 schedu]# kubectl apply -f pod3.yaml 
pod/node-affinity created
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
node-affinity              1/1     Running   0          16s   10.244.5.104   server3   <none>           <none>

部署在server3

节点亲和

指定pod node-affinity在server4上部署

[root@server2 schedu]# kubectl delete -f pod3.yaml 
pod "node-affinity" deleted
[root@server2 schedu]# vim pod3.yaml 

 15                operator: In

[root@server2 schedu]# kubectl apply -f pod3.yaml 
pod/node-affinity created
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
node-affinity              1/1     Running   0          3s    10.244.6.30    server4   <none>           <none>

pod亲和与反亲和

• podAffinity 主要解决POD可以和哪些POD部署在同一个拓扑域中的问题
(拓扑域用主机标签实现,可以是单个主机,也可以是多个主机组成的
cluster、zone等。)
• podAntiAffinity主要解决POD不能和哪些POD部署在同一个拓扑域中的问
题。它们处理的是Kubernetes集群内部POD和POD之间的关系。
• Pod 间亲和与反亲和在与更高级别的集合(例如 ReplicaSets,
StatefulSets,Deployments 等)一起使用时,它们可能更加有用。可以
轻松配置一组应位于相同定义拓扑(例如,节点)中的工作负载。

pod亲和

设置nginx与mysql为pod亲和，即二者需要部署在同一节点上

[root@server2 schedu]# kubectl delete -f pod3.yaml 
pod "node-affinity" deleted
[root@server2 schedu]# vim pod4.yaml
[root@server2 schedu]# cat pod4.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  containers:
  - name: nginx
    image: nginx
---
apiVersion: v1
kind: Pod
metadata:
  name: mysql
  labels:
    app: mysql
spec:
  containers:
  - name: mysql
    image: mysql
    env:
    - name: "MYSQL_ROOT_PASSWORD"
      value: "westos"
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - key: app
            operator: In
            values:
            - nginx
        topologyKey: kubernetes.io/hostname

[root@server2 schedu]# kubectl apply -f pod4.yaml 
pod/nginx created
pod/mysql created
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS         RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
mysql                      1/1     Running        0          26s   10.244.5.106   server3   <none>           <none>
nginx                      1/1     Running        0          26s   10.244.5.105   server3   <none>           <none>

pod反亲和

设置nginx与mysql为pod反亲和，即二者需要部署在不同节点上

[root@server2 schedu]# kubectl delete -f pod4.yaml 
pod "nginx" deleted
pod "mysql" deleted
[root@server2 schedu]# vim pod4.yaml 

 26     podAntiAffinity:

[root@server2 schedu]# kubectl apply -f pod4.yaml 
pod/nginx created
pod/mysql created
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS              RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
mysql                      1/1     Running             0          2s    <none>         server4   10.244.5.108            <none>
nginx                      1/1     Running             0          2s    10.244.5.107   server3   <none>           <none>

Taints(污点)

• NodeAffinity节点亲和性,是Pod上定义的一种属性,使Pod能够按我们的要求调度
到某个Node上,而Taints则恰恰相反,它可以让Node拒绝运行Pod,甚至驱逐Pod。
• Taints(污点)是Node的一个属性,设置了Taints后,所以Kubernetes是不会将Pod
调度到这个Node上的,于是Kubernetes就给Pod设置了个属性Tolerations(容忍),
只要Pod能够容忍Node上的污点,那么Kubernetes就会忽略Node上的污点,就能
够(不是必须)把Pod调度过去。

Nodename可以无视任何污点

[root@server2 schedu]# cat pod1.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
spec:
  containers:

  - name: nginx
    image: nginx
  nodeName: server2

[root@server2 schedu]# kubectl apply -f pod1.yml 
pod/nginx created

server2的污点

[root@server2 schedu]# kubectl  describe  nodes server2 | grep Taints
Taints:             node-role.kubernetes.io/master:NoSchedule

已经部署到server2上了，说明nodename可以掩盖污点

[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      1/1     Running   0          41s   10.244.0.21    server2   <none>           <none>

标签方式选择node

[root@server2 schedu]# kubectl delete -f pod1.yaml 
pod "nginx" deleted
[root@server2 schedu]# cat pod2.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:
  - name: nginx
    image: nginx
  nodeSelector:
    roles: master
[root@server2 schedu]# kubectl apply -f pod2.yaml 
pod/nginx created
[root@server2 schedu]# kubectl label nodes server2 roles=master
node/server2 labeled

查看server2上的标签roles=master

[root@server2 schedu]# kubectl get node --show-labels 
NAME      STATUS   ROLES                  AGE   VERSION   LABELS
server2   Ready    control-plane,master   11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=server2,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=,node.kubernetes.io/exclude-from-external-load-balancers=,roles=master
server3   Ready    <none>                 11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=server3,kubernetes.io/os=linux
server4   Ready    <none>                 11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=server4,kubernetes.io/os=linux

查看pod 处于pending状态，说明污点优先级高于标签选择

[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      0/1     Pending   0          73s   <none>         <none>    <none>           <none>

删除标签

[root@server2 schedu]# kubectl label nodes server2 roles-
node/server2 not labeled
[root@server2 schedu]# kubectl get node --show-labels 
NAME      STATUS   ROLES                  AGE   VERSION   LABELS
server2   Ready    control-plane,master   11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=server2,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=,node.kubernetes.io/exclude-from-external-load-balancers=
server3   Ready    <none>                 11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=server3,kubernetes.io/os=linux
server4   Ready    <none>                 11d   v1.21.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=server4,kubernetes.io/os=linux

容忍标签

tolerations中定义的key、value、effect,要与node上设置的taint保持一直:
• 如果 operator 是 Exists ,value可以省略。
• 如果 operator 是 Equal ,则key与value之间的关系必须相等。
• 如果不指定operator属性,则默认值为Equal。
还有两个特殊值:
• 当不指定key,再配合Exists 就能匹配所有的key与value ,可以容忍所有污点。
• 当不指定effect ,则匹配所有的effect。

NoSchedule

设置sevrer3 taint key=value:NoSchedule

[root@server2 schedu]# kubectl  taint nodes server3 key=value:NoSchedule
node/server3 tainted
[root@server2 schedu]# cat pod2.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:
  - name: nginx
    image: nginx
[root@server2 schedu]# kubectl apply -f pod2.yaml 
pod/nginx created
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      1/1     Running   0          7s    10.244.6.32    server4   <none>           <none>

server3上有污点部署在server4

为server3加标签roles=master

[root@server2 schedu]# vim pod2.yaml 
[root@server2 schedu]# cat pod2.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:
  - name: nginx
    image: nginx
  nodeSelector:
    roles: master
  #tolerations:
  #- operator: "Exists"
  #  effect: "NoSchedule"
[root@server2 schedu]# kubectl label nodes server3 roles=master
node/server3 labeled

查看pod pending

[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      0/1     Pending   0          37s   <none>         server3   <none>           <none>

为pod添加容忍 running

[root@server2 schedu]# cat pod2.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:
  - name: nginx
    image: nginx
  nodeSelector:
    roles: master
  tolerations:
  - operator: "Exists"
    effect: "NoSchedule"
[root@server2 schedu]# kubectl apply -f pod2.yaml 
pod/nginx configured
[root@server2 schedu]# kubectl get pod -o wide
NAME                       READY   STATUS    RESTARTS   AGE   IP             NODE      NOMINATED NODE   READINESS GATES
nginx                      1/1     Running   0          57s   10.244.5.108   server3   <none>           <none>

NoExecute

删除污点

[root@server2 schedu]# kubectl taint nodes server3 key:NoSchedule-
node/server3 untainted

添加容器

[root@server2 schedu]# kubectl apply -f deployment.yaml 
deployment.apps/nginx-deployment created
deployment.apps/myapp-deployment created
[root@server2 pod]# cat deployment.yaml 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
  labels:
    app: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: myapp:v1
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: myapp-deployment
  labels:
    app: myapp
spec:
  replicas: 3
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      labels:
        app: myapp
    spec:
      containers:
      - name: myapp
        image: myapp:v2

[root@server2 schedu]# kubectl get pod -o wide
NAME                                READY   STATUS    RESTARTS   AGE     IP             NODE      NOMINATED NODE   READINESS GATES
myapp-deployment-67f8c948cf-6mzm5   1/1     Running   0          42s     10.244.5.113   server3   <none>           <none>
myapp-deployment-67f8c948cf-bg7fl   1/1     Running   0          42s     10.244.5.110   server3   <none>           <none>
myapp-deployment-67f8c948cf-s7m5b   1/1     Running   0          42s     10.244.5.111   server3   <none>           <none>
mychart-6675bd6ffd-dh8hs            1/1     Running   1          22h     10.244.5.101   server3   <none>           <none>
nginx                               1/1     Running   0          9m12s   10.244.5.109   server3   <none>           <none>
nginx-deployment-6456d7c676-h88rx   1/1     Running   0          42s     10.244.5.114   server3   <none>           <none>
nginx-deployment-6456d7c676-ktkdv   1/1     Running   0          42s     10.244.5.115   server3   <none>           <none>
nginx-deployment-6456d7c676-w5gkt   1/1     Running   0          42s     10.244.5.112   server3   <none>           <none>

看到容器全在server3上给server3加污点

[root@server2 schedu]# kubectl taint node  server3 key1=v1:NoExecute
node/server3 tainted
[root@server2 schedu]# kubectl get pod -o wide
NAME                                READY   STATUS              RESTARTS   AGE    IP             NODE      NOMINATED NODE   READINESS GATES
[root@server2 schedu]# kubectl get pod -o wide
NAME                                READY   STATUS              RESTARTS   AGE   IP       NODE      NOMINATED NODE   READINESS GATES
myapp-deployment-67f8c948cf-pn8g5   0/1     ContainerCreating   0          24s   <none>   server4   <none>           <none>
myapp-deployment-67f8c948cf-vbbsr   0/1     ContainerCreating   0          20s   <none>   server4   <none>           <none>
myapp-deployment-67f8c948cf-xkp99   0/1     ContainerCreating   0          24s   <none>   server4   <none>           <none>
mychart-6675bd6ffd-hcs56            0/1     ContainerCreating   0          24s   <none>   server4   <none>           <none>
nginx-deployment-6456d7c676-6fsgx   0/1     ContainerCreating   0          24s   <none>   server4   <none>           <none>
nginx-deployment-6456d7c676-dg6fx   0/1     ContainerCreating   0          22s   <none>   server4   <none>           <none>
nginx-deployment-6456d7c676-ppz59   0/1     ContainerCreating   0          24s   <none>   server4   <none>           <none>

看到容器被驱逐server4上

添加容忍标签之后server3 又可以运行容器了

[root@server2 schedu]# cat pod1.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: nginx
spec:
  containers:

  - name: nginx
    image: nginx
  nodeName: server3
  tolerations:
   - key: "key1"
     operator: "Equal"
     value: "v1"
     effect: "NoExecute"
nginx                               1/1     Running       0          2s    10.244.5.116   server3   <none>           <none>

cordon、drain、delete

• 影响Pod调度的指令还有:cordon、drain、delete,后期创建的pod都不会被调度
到该节点上,但操作的暴力程度不一样。

cordon 停止调度

影响最小，只会将node调为SchedulingDisabled，新创建pod，不会被调度到该节点，节点原有pod不受影响，仍正常对外提供服务。

[root@server2 schedu]# kubectl cordon server3
node/server3 cordoned
[root@server2 schedu]# kubectl get node
NAME      STATUS                     ROLES                  AGE   VERSION
server2   Ready                      control-plane,master   11d   v1.21.3
server3   Ready,SchedulingDisabled   <none>                 11d   v1.21.3
server4   Ready                      <none>                 11d   v1.21.3

恢复

[root@server2 schedu]# kubectl uncordon server3
node/server3 uncordoned
[root@server2 schedu]# kubectl get node
NAME      STATUS   ROLES                  AGE   VERSION
server2   Ready    control-plane,master   11d   v1.21.3
server3   Ready    <none>                 11d   v1.21.3
server4   Ready    <none>                 11d   v1.21.3

drain 驱逐节点

首先驱逐node上的pod，在其他节点重新创建，然后将节点调为SchedulingDisabled。

[root@server2 schedu]# kubectl drain server3
node/server3 cordoned
evicting pod "web-1"
evicting pod "coredns-9d85f5447-mgg2k"
pod/coredns-9d85f5447-mgg2k evicted
pod/web-1 evicted
node/server3 evicted

恢复

[root@server2 schedu]# kubectl uncordon server3
node/server3 uncordoned

delete 删除节点

最暴力的一个，首先驱逐node上的pod，在其他节点重新创建，然后，从master节点删除该node，master失去对其控制，如要恢复调度，需进入node节点，重启kubelet服务

kubectl delete node server3

systemctl restart kubelet //基于node的自注册功能,恢复使用

oam_chen

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
企业运维----Docker-kubernetes-调度

kubernetes-调度 k8s调度概述nodename节点选择约束nodeSelector节点选择约束亲和与反亲和节点亲和节点反亲和节点亲和pod亲和与反亲和pod亲和pod反亲和 Taints(污点)Nodename可以无视任何污点标签方式选择node容忍标签NoScheduleNoExecutecordon、drain、deletecordon 停止调度drain 驱逐节点delete 删除节点 k8s调度概述• 调度器通过 kubernetes 的 watch 机制来发现集群中新创建且尚未
复制链接

扫一扫

专栏目录