Kubernetes K8S之固定节点nodeName和nodeSelector调度详解当nodeName指定节点存在当nodeName指定节点不存在添加label

踏歌行

发布于 2020-12-07 11:26:55

5.1K0

发布于 2020-12-07 11:26:55

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。