Kubernetes K8S之固定节点nodeName和nodeSelector调度详解与示例

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

nodeName调度

nodeName是节点选择约束的最简单形式，但是由于其限制，通常很少使用它。nodeName是PodSpec的领域。

pod.spec.nodeName将Pod直接调度到指定的Node节点上，会【跳过Scheduler的调度策略】，该匹配规则是【强制】匹配。可以越过Taints污点进行调度。

nodeName用于选择节点的一些限制是：

  • 如果指定的节点不存在，则容器将不会运行，并且在某些情况下可能会自动删除。
  • 如果指定的节点没有足够的资源来容纳该Pod，则该Pod将会失败，并且其原因将被指出，例如OutOfmemory或OutOfcpu。
  • 云环境中的节点名称并非总是可预测或稳定的。

nodeName示例

获取当前的节点信息

1 [root@k8s-master scheduler]# kubectl get nodes -o wide

2 NAME         STATUS   ROLES    AGE   VERSION   INTERNAL-IP    EXTERNAL-IP   OS-IMAGE                KERNEL-VERSION           CONTAINER-RUNTIME

3 k8s-master   Ready    master   42d   v1.17.4   172.16.1.110   <none>        CentOS Linux 7 (Core)   3.10.0-1062.el7.x86_64   docker://19.3.8

4 k8s-node01   Ready    <none>   42d   v1.17.4   172.16.1.111   <none>        CentOS Linux 7 (Core)   3.10.0-1062.el7.x86_64   docker://19.3.8

5 k8s-node02   Ready    <none>   42d   v1.17.4   172.16.1.112   <none>        CentOS Linux 7 (Core)   3.10.0-1062.el7.x86_64   docker://19.3.8

当nodeName指定节点存在

要运行的yaml文件

 1 [root@k8s-master scheduler]# pwd

 2 /root/k8s_practice/scheduler

 3 [root@k8s-master scheduler]# cat scheduler_nodeName.yaml

 4 apiVersion: apps/v1

 5 kind: Deployment

 6 metadata:

 7   name: scheduler-nodename-deploy

 8   labels:

 9     app: nodename-deploy

10 spec:

11   replicas: 5

12   selector:

13     matchLabels:

14       app: myapp

15   template:

16     metadata:

17       labels:

18         app: myapp

19     spec:

20       containers:

21       - name: myapp-pod

22         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1

23         imagePullPolicy: IfNotPresent

24         ports:

25           - containerPort: 80

26       # 指定节点运行

27       nodeName: k8s-master

运行yaml文件并查看信息

 1 [root@k8s-master scheduler]# kubectl apply -f scheduler_nodeName.yaml

 2 deployment.apps/scheduler-nodename-deploy created

 3 [root@k8s-master scheduler]#

 4 [root@k8s-master scheduler]# kubectl get deploy -o wide

 5 NAME                        READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR

 6 scheduler-nodename-deploy   0/5     5            0           6s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp

 7 [root@k8s-master scheduler]#

 8 [root@k8s-master scheduler]# kubectl get rs -o wide

 9 NAME                                  DESIRED   CURRENT   READY   AGE   CONTAINERS   IMAGES                                                      SELECTOR

10 scheduler-nodename-deploy-d5c9574bd   5         5         5       15s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp,pod-template-hash=d5c9574bd

11 [root@k8s-master scheduler]#

12 [root@k8s-master scheduler]# kubectl get pod -o wide

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

14 scheduler-nodename-deploy-d5c9574bd-6l9d8   1/1     Running   0          23s   10.244.0.123   k8s-master   <none>           <none>

15 scheduler-nodename-deploy-d5c9574bd-c82cc   1/1     Running   0          23s   10.244.0.119   k8s-master   <none>           <none>

16 scheduler-nodename-deploy-d5c9574bd-dkkjg   1/1     Running   0          23s   10.244.0.122   k8s-master   <none>           <none>

17 scheduler-nodename-deploy-d5c9574bd-hcn77   1/1     Running   0          23s   10.244.0.121   k8s-master   <none>           <none>

18 scheduler-nodename-deploy-d5c9574bd-zstjx   1/1     Running   0          23s   10.244.0.120   k8s-master   <none>           <none>

由上可见，yaml文件中nodeName: k8s-master生效，所有pod被调度到了k8s-master节点。如果这里是nodeName: k8s-node02，那么就会直接调度到k8s-node02节点。

当nodeName指定节点不存在

要运行的yaml文件

 1 [root@k8s-master scheduler]# pwd

 2 /root/k8s_practice/scheduler

 3 [root@k8s-master scheduler]# cat scheduler_nodeName_02.yaml

 4 apiVersion: apps/v1

 5 kind: Deployment

 6 metadata:

 7   name: scheduler-nodename-deploy

 8   labels:

 9     app: nodename-deploy

10 spec:

11   replicas: 5

12   selector:

13     matchLabels:

14       app: myapp

15   template:

16     metadata:

17       labels:

18         app: myapp

19     spec:

20       containers:

21       - name: myapp-pod

22         image: registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1

23         imagePullPolicy: IfNotPresent

24         ports:

25           - containerPort: 80

26       # 指定节点运行，该节点不存在

27       nodeName: k8s-node08

运行yaml文件并查看信息

 1 [root@k8s-master scheduler]# kubectl apply -f scheduler_nodeName_02.yaml

 2 deployment.apps/scheduler-nodename-deploy created

 3 [root@k8s-master scheduler]#

 4 [root@k8s-master scheduler]# kubectl get deploy -o wide

 5 NAME                        READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR

 6 scheduler-nodename-deploy   0/5     5            0           4s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp

 7 [root@k8s-master scheduler]#

 8 [root@k8s-master scheduler]# kubectl get rs -o wide

 9 NAME                                   DESIRED   CURRENT   READY   AGE   CONTAINERS   IMAGES                                                      SELECTOR

10 scheduler-nodename-deploy-75944bdc5d   5         5         0       9s    myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp,pod-template-hash=75944bdc5d

11 [root@k8s-master scheduler]#

12 [root@k8s-master scheduler]# kubectl get pod -o wide

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

14 scheduler-nodename-deploy-75944bdc5d-c8f5d   0/1     Pending   0          13s   <none>   k8s-node08   <none>           <none>

15 scheduler-nodename-deploy-75944bdc5d-hfdlv   0/1     Pending   0          13s   <none>   k8s-node08   <none>           <none>

16 scheduler-nodename-deploy-75944bdc5d-q9qgt   0/1     Pending   0          13s   <none>   k8s-node08   <none>           <none>

17 scheduler-nodename-deploy-75944bdc5d-q9zl7   0/1     Pending   0          13s   <none>   k8s-node08   <none>           <none>

18 scheduler-nodename-deploy-75944bdc5d-wxsnv   0/1     Pending   0          13s   <none>   k8s-node08   <none>           <none>

由上可见，如果指定的节点不存在，则容器将不会运行，一直处于Pending 状态。

nodeSelector调度

nodeSelector是节点选择约束的最简单推荐形式。nodeSelector是PodSpec的领域。它指定键值对的映射。

Pod.spec.nodeSelector是通过Kubernetes的label-selector机制选择节点，由调度器调度策略匹配label，而后调度Pod到目标节点，该匹配规则属于【强制】约束。由于是调度器调度，因此不能越过Taints污点进行调度。

nodeSelector示例

获取当前的节点信息

1 [root@k8s-master ~]# kubectl get node -o wide --show-labels

2 NAME         STATUS   ROLES    AGE   VERSION   INTERNAL-IP    EXTERNAL-IP   OS-IMAGE                KERNEL-VERSION           CONTAINER-RUNTIME   LABELS

3 k8s-master   Ready    master   42d   v1.17.4   172.16.1.110   <none>        CentOS Linux 7 (Core)   3.10.0-1062.el7.x86_64   docker://19.3.8     beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master,kubernetes.io/os=linux,node-role.kubernetes.io/master=

4 k8s-node01   Ready    <none>   42d   v1.17.4   172.16.1.111   <none>        CentOS Linux 7 (Core)   3.10.0-1062.el7.x86_64   docker://19.3.8     beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01,kubernetes.io/os=linux

5 k8s-node02   Ready    <none>   42d   v1.17.4   172.16.1.112   <none>        CentOS Linux 7 (Core)   3.10.0-1062.el7.x86_64   docker://19.3.8     beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02,kubernetes.io/os=linux

添加label标签

运行kubectl get nodes以获取群集节点的名称。然后可以对指定节点添加标签。比如：k8s-node01的磁盘为SSD，那么添加disk-type=ssd；k8s-node02的CPU核数高，那么添加cpu-type=hight；如果为Web机器，那么添加service-type=web。怎么添加标签可以根据实际规划情况而定。

 1 ### 给k8s-node01 添加指定标签

 2 [root@k8s-master ~]# kubectl label nodes k8s-node01 disk-type=ssd

 3 node/k8s-node01 labeled

 4 #### 删除标签命令 kubectl label nodes k8s-node01 disk-type-

 5 [root@k8s-master ~]#

 6 [root@k8s-master ~]# kubectl get node --show-labels

 7 NAME         STATUS   ROLES    AGE   VERSION   LABELS

 8 k8s-master   Ready    master   42d   v1.17.4   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master,kubernetes.io/os=linux,node-role.kubernetes.io/master=

 9 k8s-node01   Ready    <none>   42d   v1.17.4   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disk-type=ssd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01,kubernetes.io/os=linux

10 k8s-node02   Ready    <none>   42d   v1.17.4   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02,kubernetes.io/os=linux

由上可见，已经为k8s-node01节点添加了disk-type=ssd 标签。

当nodeSelector标签存在

要运行的yaml文件

 1 [root@k8s-master scheduler]# pwd

 2 /root/k8s_practice/scheduler

 3 [root@k8s-master scheduler]#

 4 [root@k8s-master scheduler]# cat scheduler_nodeSelector.yaml

 5 apiVersion: apps/v1

 6 kind: Deployment

 7 metadata:

 8   name: scheduler-nodeselector-deploy

 9   labels:

10     app: nodeselector-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

27       # 指定节点标签选择，且标签存在

28       nodeSelector:

29         disk-type: ssd

运行yaml文件并查看信息

 1 [root@k8s-master scheduler]# kubectl apply -f scheduler_nodeSelector.yaml

 2 deployment.apps/scheduler-nodeselector-deploy created

 3 [root@k8s-master scheduler]#

 4 [root@k8s-master scheduler]# kubectl get deploy -o wide

 5 NAME                            READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR

 6 scheduler-nodeselector-deploy   5/5     5            5           10s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp

 7 [root@k8s-master scheduler]#

 8 [root@k8s-master scheduler]# kubectl get rs -o wide

 9 NAME                                       DESIRED   CURRENT   READY   AGE   CONTAINERS   IMAGES                                                      SELECTOR

10 scheduler-nodeselector-deploy-79455db454   5         5         5       14s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp,pod-template-hash=79455db454

11 [root@k8s-master scheduler]#

12 [root@k8s-master scheduler]# kubectl get pod -o wide

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

14 scheduler-nodeselector-deploy-79455db454-745ph   1/1     Running   0          19s   10.244.4.154   k8s-node01   <none>           <none>

15 scheduler-nodeselector-deploy-79455db454-bmjvd   1/1     Running   0          19s   10.244.4.151   k8s-node01   <none>           <none>

16 scheduler-nodeselector-deploy-79455db454-g5cg2   1/1     Running   0          19s   10.244.4.153   k8s-node01   <none>           <none>

17 scheduler-nodeselector-deploy-79455db454-hw8jv   1/1     Running   0          19s   10.244.4.152   k8s-node01   <none>           <none>

18 scheduler-nodeselector-deploy-79455db454-zrt8d   1/1     Running   0          19s   10.244.4.155   k8s-node01   <none>           <none>

由上可见，所有pod都被调度到了k8s-node01节点。当然如果其他节点也有disk-type=ssd 标签，那么pod也会调度到这些节点上。

当nodeSelector标签不存在

要运行的yaml文件

 1 [root@k8s-master scheduler]# pwd

 2 /root/k8s_practice/scheduler

 3 [root@k8s-master scheduler]#

 4 [root@k8s-master scheduler]# cat scheduler_nodeSelector_02.yaml

 5 apiVersion: apps/v1

 6 kind: Deployment

 7 metadata:

 8   name: scheduler-nodeselector-deploy

 9   labels:

10     app: nodeselector-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

27       # 指定节点标签选择，且标签不存在

28       nodeSelector:

29         service-type: web

运行yaml文件并查看信息

 1 [root@k8s-master scheduler]# kubectl apply -f scheduler_nodeSelector_02.yaml

 2 deployment.apps/scheduler-nodeselector-deploy created

 3 [root@k8s-master scheduler]#

 4 [root@k8s-master scheduler]# kubectl get deploy -o wide

 5 NAME                            READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                                      SELECTOR

 6 scheduler-nodeselector-deploy   0/5     5            0           26s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp

 7 [root@k8s-master scheduler]#

 8 [root@k8s-master scheduler]# kubectl get rs -o wide

 9 NAME                                       DESIRED   CURRENT   READY   AGE   CONTAINERS   IMAGES                                                      SELECTOR

10 scheduler-nodeselector-deploy-799d748db6   5         5         0       30s   myapp-pod    registry.cn-beijing.aliyuncs.com/google_registry/myapp:v1   app=myapp,pod-template-hash=799d748db6

11 [root@k8s-master scheduler]#

12 [root@k8s-master scheduler]# kubectl get pod -o wide

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

14 scheduler-nodeselector-deploy-799d748db6-92mqj   0/1     Pending   0          40s   <none>   <none>   <none>           <none>

15 scheduler-nodeselector-deploy-799d748db6-c2w25   0/1     Pending   0          40s   <none>   <none>   <none>           <none>

16 scheduler-nodeselector-deploy-799d748db6-c8tlx   0/1     Pending   0          40s   <none>   <none>   <none>           <none>

17 scheduler-nodeselector-deploy-799d748db6-tc5n7   0/1     Pending   0          40s   <none>   <none>   <none>           <none>

18 scheduler-nodeselector-deploy-799d748db6-z8c57   0/1     Pending   0          40s   <none>   <none>   <none>           <none>

由上可见，如果nodeSelector匹配的标签不存在，则容器将不会运行，一直处于Pending 状态。

相关阅读

1、官网：Pod分配调度

2、Kubernetes K8S之调度器kube-scheduler详解

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

4、Kubernetes K8S之Taints污点与Tolerations容忍详解

完毕！

———END———
如果觉得不错就关注下呗 (-^O^-) ！

Kubernetes K8S之固定节点nodeName和nodeSelector调度详解的更多相关文章

  1. Kubernetes K8S之通过helm部署metrics-server与HPA详解

    Kubernetes K8S之通过helm部署metrics-server与 Horizontal Pod Autoscaling (HPA)详解 主机配置规划 服务器名称(hostname) 系统版 ...

  2. Kubernetes服务pod的健康检测liveness和readiness详解

    Kubernetes服务pod的健康检测liveness和readiness详解 接下来给大家讲解下在K8S上,我们如果对我们的业务服务进行健康检测. Health Check.restartPoli ...

  3. kubernetes 亲和性调度详解

    文章目录 1 概述: 2 场景一:调度到一组具有相同特性的主机上(label+nodeSelector) 3 场景二:部署的应用不想调度到某些节点上(nodeaffinity) 4 场景三:部署的应用 ...

  4. k8s入坑之路（7）kubernetes设计精髓List/Watch机制和Informer模块详解

    1.list-watch是什么 List-watch 是 K8S 统一的异步消息处理机制,保证了消息的实时性,可靠性,顺序性,性能等等,为声明式风格的API 奠定了良好的基础,它是优雅的通信方式,是 ...

  5. 资深实践篇 | 基于Kubernetes 1.61的Kubernetes Scheduler 调度详解

    欢迎大家前往腾讯云技术社区,获取更多腾讯海量技术实践干货哦~ 作者:腾讯云容器服务团队 源码为 k8s v1.6.1 版本,github 上对应的 commit id 为 b0b7a323cc5a4a ...

  6. K8s 从懵圈到熟练 – 集群网络详解

    作者 | 声东 阿里云售后技术专家 导读:阿里云 K8S 集群网络目前有两种方案:一种是 flannel 方案:另外一种是基于 calico 和弹性网卡 eni 的 terway 方案.Terway ...

  7. ROS节点的初始化及退出详解（ros::init、SIGINT、ros::ok、ros::NodeHandle

    https://haoqchen.site/2018/04/28/ROS-node-init/ #include "ros/ros.h" #include <signal.h ...

  8. Kubernetes K8S之Ingress详解与示例

    K8S之Ingress概述与说明,并详解Ingress常用示例 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 ...

  9. Kubernetes K8S之affinity亲和性与反亲和性详解与示例

    Kubernetes K8S之Node节点亲和性与反亲和性以及Pod亲和性与反亲和性详解与示例 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-mas ...

随机推荐

  1. 基于SLF4J的MDC机制和Dubbo的Filter机制，实现分布式系统的日志全链路追踪

    原文链接:基于SLF4J的MDC机制和Dubbo的Filter机制,实现分布式系统的日志全链路追踪 一.日志系统 1.日志框架 在每个系统应用中,我们都会使用日志系统,主要是为了记录必要的信息和方便排 ...

  2. VSCode--HTML代码片段（基础版，react、vue、jquery）

    起因是最近在学习前端,看的网上的demo也是在react.vue.jquery之间穿插,为了方便一键生成html模板(懒)写demo,有了以下折腾. 本人使用的前端编辑工具是vscode(方便.懒), ...

  3. 剑指29：最小的k个数

    题目描述 输入n个整数,找出其中最小的K个数.例如输入4,5,1,6,2,7,3,8这8个数字,则最小的4个数字是1,2,3,4. class Solution {public:    vector& ...

  4. 1 Prism概述

    架构目标 以模块化方式开发应用,这些模块被独立团队用WPF技术开发,集成,部署,这是使用Prism的最大好处. 最小化交叉团队依赖.允许团队在不同领域专业化,比如UI设计,商业逻辑实现,基础代码开发 ...

  5. 3.java设计模式之工厂模式

    基本需求: 一个披萨店需要订购不同种类的披萨 传统方式: 实现思路 在订购类中根据用户不同的输入直接创建不同的披萨实体类进行返回 UML类图 代码实现 披萨类 // 抽象父类 public abstr ...

  6. NIO源码分析：SelectionKey

    SelectionKey SelectionKey,选择键,在每次通道注册到选择器上时都会创建一个SelectionKey储存在该选择器上,该SelectionKey保存了注册的通道.注册的选择器.通 ...

  7. SpringBoot中的响应式web应用

    目录 简介 Reactive in Spring 注解方式使用WebFlux 编程方式使用webFlux Spring WebFlux的测试 总结 简介 在Spring 5中,Spring MVC引入 ...

  8. MySQL慢查询开启、日志分析(转)

    说明 Mysql的查询讯日志是Mysql提供的一种日志记录,它用来记录在Mysql中响应时间超过阈值的语句 具体指运行时间超过long_query_time值得SQL,则会被记录到慢查询日志中.lon ...

  9. 【webSokect】基于django Channels的简单实现

    # settings.py: INSTALLED_APPS = [ 'django.contrib.admin', 'django.contrib.auth', 'django.contrib.con ...

  10. MTK官方SDK包编译openwrt

    全过程需要联网,最好有梯子,编译方式有两种 安装依赖库: apt-get install g++ apt-get install libncurses5-dev apt-get install zli ...