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

 

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. 172.16   。172.16 。 110     <无> CentOS Linux (Core)     3。el7.x 11062 .el7.x 1064 docker: // 19.3 - 4 k8s-node01 4.7.8 4 k8s-node01 4.8.7 <none> 22d v1. 17.4 172.16。1.111     <无> CentOS Linux 7（内核）    3.10 . 0 - 1062
    .el7.x86_64 docker: // 19.3.8 
5 k8s-node02 就绪 <none> 42d v1. 17.4    172.16。1.112    <无> CentOS Linux 7 (Core)    3.10。0 - 1062 .el7.x86_64 泊坞窗：// 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-节点名-部署
8个   标签：
 9     应用程序：节点名-部署
10 规范：
 11   副本：5 
12   选择器：
 13     匹配标签：
 14       应用程序：myapp
 15   模板：
 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 节点 NOMINATED 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     运行   0           23s    10.244。0.122    k8s-master <none> <none>
 17 scheduler-nodename-deploy-d5c9574bd-hcn77    1 / 1     运行   0           23s    10.244。0.121    k8s-master <none> <none>
 18 scheduler-nodename-deploy-d5c9574bd-zstjx    1 / 1     运行   0           23s    10.244。0.120    k8s-master <无> <无>

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-节点名-部署
8个   标签：
 9     应用程序：节点名-部署
10 规范：
 11   副本：5 
12   选择器：
 13     匹配标签：
 14       应用程序：myapp
 15   模板：
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宽
13 NAME READY STATUS RESTARTS AGE IP Node NOMINATED 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> <
无>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 <无> <无>

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    <无> CentOS Linux 7 (Core)    3.10。0 - 1062 .el7.x86_64 docker: // 19.3.8 beta./arch=amd64,beta./os=linux,/arch=amd64,/hostname=k8s-大师，/os=linux，node-role./master= 
4k8s-node01 就绪 <无> 42d v1. 17.4    172.16。1.111    <无> CentOS Linux 7 (Core)    3.10。0 - 1062 .el7.x86_64 docker: // 19.3.8 beta./arch=amd64,beta./os=linux,/arch=amd64,/hostname=k8s- node01,/os=linux 
5 k8s-node02 Ready <none> 42d v1. 17.4    172.16。1.112    <无> CentOS Linux 7 (Core)    3.10。0 - 1062 .el7.x86_64 泊坞窗：//19.3.8     beta./arch=amd64,beta./os=linux,/arch=amd64,/hostname=k8s-node02,/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./arch=amd64,beta./os=linux,/arch=amd64,/hostname=k8s-master,/os=linux,node-role./master=
 9 k8s-node01 Ready <none> 42d v1. 17.4    beta./arch=amd64,beta./os=linux,disk-type=ssd,/arch=amd64,/主机名=k8s-node01,/os = linux
 10 k8s-node02 就绪 <无> 42d v1。17.4    beta./arch=amd64,beta./os=linux,/arch=amd64,/主机名=k8s-node02,/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种 ：部署
7 元数据：
 8   名称：scheduler-nodeselector-部署
9   标签：
 10     应用程序：nodeselector-部署
11 规范：
 12   副本：5 
13   选择器：
 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-主调度器]# 
  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宽
13 名称就绪状态重新启动年龄 IP 节点提名节点就绪门
14 scheduler-nodeselector-deploy-79455db454-745ph    1 / 1     运行   0           19s    10.244。4.154    k8s-node01 <none> <none>
 15 scheduler-nodeselector-deploy-79455db454-bmjvd    1 / 1     运行   0           19s    10.244。4.151    k8s-node01 <none> <none>
 16调度程序-nodeselector-deploy-79455db454-g5cg2    1 /1     运行   0           19s    10.244。4.153 k8s    -node01 <none> <none>
 17 scheduler-nodeselector-deploy-79455db454-hw8jv    1 / 1     运行   0           19s    10.244。4.152    k8s-node01 <none> <none>
 18 scheduler-nodeselector-deploy-79455db454-zrt8d    1 / 1     运行   0           19s    10.244。4.155    k8s-node01 <无> <无>

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种 ：部署
7 元数据：
 8   名称：scheduler-nodeselector-部署
9   标签：
 10     应用程序：nodeselector-部署
11 规范：
 12   副本：5 
13   选择器：
 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 -主调度程序]# 
  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宽
13  NAME READY STATUS RESTARTS AGE IP Node NOMINATED节点就绪
 门14 scheduler-nodeselector-deploy-799d748db6-92mqj     0 / 1      待定    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      等待    0            40s <无> <无> <无>
<无> 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> <无> <无> <无>

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。