随笔- 79 文章- 0 评论- 2 阅读- 35619

每天一点基础K8S--K8S中调度策略---节点选择器nodeName、nodeSelector

K8S中的调度策略---节点选择器

正常情况下，Pod被调度到哪一个节点是通过scheduler完成，但在一些场景中，时常需要将一些pod指定在某一个node上运行。那此时就可以通过nodeName和nodeSelector完成

nodeName

   nodeName	<string>
     NodeName is a request to schedule this pod onto a specific node. If it is non-empty, the scheduler simply schedules this pod onto that node, assuming that it fits resource requirements.

[root@master-worker-node-1 pod]# cat test-node-name.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: test-node-name
  labels: 
    function: test-node-name
spec:
  nodeName: only-worker-node-3
  containers:
  - name: test-ndoe-name-container
    image: busybxox:latest
    imagePullPolicy: IfNotPresent
    command: ['/bin/sh','-c','sleep 1234']

创建pod
[root@master-worker-node-1 pod]# kubectl apply -f test-node-name.yaml 
pod/test-node-name created

pod被调度到了only-worker-node-3
[root@master-worker-node-1 pod]# kubectl get pods --show-labels -o wide
NAME             READY   STATUS    RESTARTS   AGE   IP            NODE                 NOMINATED NODE   READINESS GATES   LABELS
test-node-name   1/1     Running   0          62s   10.244.31.7   only-worker-node-3   <none>           <none>            function=test-node-name

nodeSelector

nodeSelector和后面要学习的亲和性都是根据node的label进行选择。

因为nodeSelector是和标签有关，先给其中一个worker node打上一个标签

[root@master-worker-node-1 pod]# kubectl label nodes only-worker-node-3 test-nodeselector=target-node
node/only-worker-node-3 labeled


[root@master-worker-node-1 pod]# kubectl get nodes --show-labels
NAME                   STATUS   ROLES           AGE     VERSION   LABELS
master-worker-node-1   Ready    control-plane   3d22h   v1.25.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=master-worker-node-1,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node.kubernetes.io/exclude-from-external-load-balancers=
master-worker-node-2   Ready    control-plane   3d21h   v1.25.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=master-worker-node-2,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node.kubernetes.io/exclude-from-external-load-balancers=
only-worker-node-3     Ready    worker          3d21h   v1.25.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=only-worker-node-3,kubernetes.io/os=linux,node-role.kubernetes.io/worker=,test-nodeselector=target-node
only-worker-node-4     Ready    worker          3d21h   v1.25.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=only-worker-node-4,kubernetes.io/os=linux,node-role.kubernetes.io/worker=

[root@master-worker-node-1 pod]# cat test-node-selector.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: test-node-selector
  labels: 
    function: test-node-selector
spec:
  nodeSelector: 
    test-nodeselector: target-node
  containers:
  - name: test-node-selector
    image: busybox:latest
    imagePullPolicy: IfNotPresent
    command: ['/bin/sh','-c','sleep 1234']

创建pod
[root@master-worker-node-1 pod]# kubectl apply -f test-node-selector.yaml 
pod/test-node-selector created

pod被调度到含有nodeslector中关联的label的node上。
[root@master-worker-node-1 pod]# kubectl get pods -o wide 
NAME                 READY   STATUS    RESTARTS   AGE   IP            NODE                 NOMINATED NODE   READINESS GATES
test-node-name       1/1     Running   0          18m   10.244.31.7   only-worker-node-3   <none>           <none>
test-node-selector   1/1     Running   0          11s   10.244.31.8   only-worker-node-3   <none>           <none>

将only-worker-node-4也打上对应的标签test-nodeselector=target-node
[root@master-worker-node-1 pod]# kubectl label nodes  only-worker-node-4 test-nodeselector=target-node
node/only-worker-node-4 labeled

重新新增一个pod，该pod具有相同的nodeSelector
[root@master-worker-node-1 pod]# kubectl apply -f test-node-selector-2.yaml 
pod/test-node-selector-2 created

此时pod被scheduler在具有test-nodeselector这个label的node之间进行调度。结果显示被调度到only-worker-node-4
[root@master-worker-node-1 ~]# kubectl get pods -o wide 
NAME                   READY   STATUS    RESTARTS      AGE   IP            NODE                 NOMINATED NODE   READINESS GATES
test-node-name         1/1     Running   1 (12m ago)   33m   10.244.31.7   only-worker-node-3   <none>           <none>
test-node-selector     1/1     Running   0             14m   10.244.31.8   only-worker-node-3   <none>           <none>
test-node-selector-2   1/1     Running   0             10m   10.244.54.7   only-worker-node-4   <none>           <none>

nodeName和nodeSelector优先级对比

测试前，现将only-worker-node-4的label去除
[root@master-worker-node-1 pod]# kubectl label nodes only-worker-node-4 test-nodeselector-
node/only-worker-node-4 unlabeled

[root@master-worker-node-1 pod]# cat compare-nodename-with-node-selector.yaml 
apiVersion: v1
kind: Pod
metadata: 
  name: compare
  labels:
    function: compare-nodeName-with-nodeSelector
spec:
  nodeName: only-worker-node-4      # nodeName选择only-worker-node-4
  nodeSelector:
    test-nodeselector: target-node  # nodeSelector选择only-worker-node-3
  containers:
  - name: compare
    image: busybox:latest
    imagePullPolicy: IfNotPresent
    command: ['/bin/sh','-c','sleep 1234']

[root@master-worker-node-1 pod]# kubectl apply -f compare-nodename-with-node-selector.yaml 
pod/compare created

发现pod不能正常运行
[root@master-worker-node-1 pod]# kubectl get pods -o wide |  grep compare
compare                0/1     NodeAffinity   0               105s   <none>        only-worker-node-4   <none>           <none>

查看pod详情发现nodeAffnity调度失败
[root@master-worker-node-1 pod]# kubectl describe pods compare | tail -4
Events:
  Type     Reason        Age    From     Message
  ----     ------        ----   ----     -------
  Warning  NodeAffinity  2m23s  kubelet  Predicate NodeAffinity failed

[root@master-worker-node-1 pod]# cat compare-nodename-with-node-selector-2.yaml 
apiVersion: v1
kind: Pod
metadata: 
  name: compare-2
  labels:
    function: compare-nodeName-with-nodeSelector
spec:
  nodeName: only-worker-node-3  # nodeName选择only-worker-node-3
  nodeSelector:
    test-nodeselector: target-node # nodeSelector也选择only-worker-node-3
  containers:
  - name: compare-2
    image: busybox:latest
    imagePullPolicy: IfNotPresent
    command: ['/bin/sh','-c','sleep 1234']
    
 [root@master-worker-node-1 pod]# kubectl apply -f compare-nodename-with-node-selector-2.yaml 
pod/compare-2 created

pod可以运行    
 [root@master-worker-node-1 pod]# kubectl get pods -o wide  |  grep compare
compare                0/1     NodeAffinity   0             8m56s   <none>        only-worker-node-4   <none>           <none>
compare-2              1/1     Running        0             17s     10.244.31.9   only-worker-node-3   <none>           <none>