pod&node选择部署策略： nodeSelector和nodeAffinity

1、登录服务器修改节点标签

查看所有节点labels：
kubectl get nodes --show-labels

添加labels：
kubectl label nodes node2 [label.key]=[label.value]

修改pod配置文件与工作节点标签对应
修改yaml配置文件：

添加nodeSelector配置，样式为[label.key]: [label.value]
nodeSelector:
  [label.key]: [label.value]
修改后重启服务
1
2
3
4
5
6
7
8
9
10
11
12
13

示例：

apiVersion: apps/v1
kind: Pod
metadata:
  name: nginx-pod
  namespace: namespace
  labels:
    app: nginx-pod
spec:
   nodeSelector:
     deploy.type: assembly_app
   containers:
     - name: nginx
       image: "nginx"
       restartPolicy: Always
1
2
3
4
5
6
7
8
9
10
11
12
13
14

拓展：
nodeSelector配置相对简单，k8s提供了另外一个pod调度配置：nodeAffinity（节点亲和），相对于nodeSelector的简单匹配他拥有更多更加个性化的配置。

    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: deploy.type
                    operator: In
                    values:
                      - yztssjdxt-test
                      - yztssjdxt

1
2
3
4
5
6
7
8
9
10
11
12

这段配置表示：该pod可以被调度到标签key为 “deploy.type” ，值为 “yztssjdxt-test” 或 “yztssjdxt” 的节点。

通过使用的语句不同，调度分成软策略(soft)和硬策略(hard)，在软策略下，如果没有满足调度条件的节点，pod会忽略这条规则，继续完成调度。而硬策略下，pod必须部署到满足条件的节点上，如果没有满足条件的节点，就不停重试。

DuringScheduling和IgnoredDuringExecution组合生成目前主要的nodeaffinity策略

软硬策略可以组合使用，如下面这段配置中pod必须被调度到标签为 “deploy.type=yztssjdxt-test” 的节点。另外，在满足该条件的节点中，优先使用具有 “server=oms” 标签的节点。

  spec:
    affinity:
      nodeAffinity:
        requiredDuringSchedulingIgnoredDuringExecution:
          nodeSelectorTerms:
            - matchExpressions:
                - key: deploy.type
                  operator: In
                  values:
                    - yztssjdxt-test
          preferredDuringSchedulingIgnoredDuringExecution:
              - weight: 1
                  preference:
                  matchExpressions:
                      - key: server
                          operator: In
                          values:
                              - oms

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

nodeAffinity中nodeSelector字段下，节点满足任何一个条件即可；但更下一级matchExpressions，节点必须同时满足所有条件才能运行pod 。

如下面这段配置需要deploy.type=yztssjdxt-test、server=oms同时满足

    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: deploy.type
                    operator: In
                    values:
                      - yztssjdxt-test
                  - key: server
                    operator: In
                    values:
                      - oms
1
2
3
4
5
6
7
8
9
10
11
12
13
14

而这段配置只需要deploy.type=yztssjdxt-test、deploy.type1=yztssjdxt满足一个即可

    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: deploy.type
                    operator: In
                    values:
                      - yztssjdxt-test
              - matchExpressions:
                  - key: deploy.type1
                    operator: In
                    values:
                      - yztssjdxt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

operator字段：pod与node的匹配逻辑，可选的操作符有：

如下方这段配置，pod将不会调度到deploy.type=yztssjdxt-test的node上

    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: deploy.type
                    operator: NotIn
                    values:
                      - yztssjdxt-test

1
2
3
4
5
6
7
8
9
10
11