无法通过Service Cluster IP与Pod通信，但Pod IP可正常连通

问题现象 #

• 使用ems Pod的IP地址直接连接gunicorn API正常：

root@nginx:/etc/nginx# curl -v http://10.244.1.155:20000/_ems/plant
*   Trying 10.244.1.155:20000...
* Connected to 10.244.1.155 (10.244.1.155) port 20000 (#0)

• 使用ClusterIP类型的ems-service连接时，出现连接拒绝错误：

root@nginx:/etc/nginx# curl -v http://ems-service.default.svc.cluster.local:20000/_ems/plant;
*   Trying 10.106.43.2:20000...
* connect to 10.106.43.2 port 20000 failed: Connection refused
* Failed to connect to ems-service.default.svc.cluster.local port 20000: Connection refused
* Closing connection 0
curl: (7) Failed to connect to ems-service.default.svc.cluster.local port 20000: Connection refused

相关资源状态 #

NAME                                                         READY   STATUS             RESTARTS         AGE
pod/ems                                                      1/1     Running            0                5m56s
pod/nginx                                                    1/1     Running            0                4h46m
NAME                       TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE
service/ems-service        ClusterIP   10.106.43.2    <none>        20000/TCP        103m

执行kubectl describe service -n default ems-service输出：

Name:              ems-service
Namespace:         default
Labels:            <none>
Annotations:       <none>
Selector:          name=ems
Type:              ClusterIP
IP Family Policy:  SingleStack
IP Families:       IPv4
IP:                10.106.43.2
IPs:               10.106.43.2
Port:              <unset>  20000/TCP
TargetPort:        80/TCP
Endpoints:         <none>
Session Affinity:  None
Events:            <none>

Pod与Service的YAML配置 #

ems Pod配置： #

apiVersion: v1
kind: Pod
metadata:
  name: ems
  labels:
    app: ems
spec:
  containers:
    - name: ems
      image: 
      command: ["/bin/sh"]
      args: ["-c", "chmod +x /root/code/run.sh && /root/code/run.sh"]
      env:
      - name: LOGS_DIR
        value: /root/code/logs
      imagePullPolicy: Always
      ports:
      - containerPort: 20000

ems-service配置： #

apiVersion: v1
kind: Service
metadata:
  name: ems-service
spec:
  selector:
    name: ems
  ports:
    - protocol: TCP
      port: 20000
      targetPort: 80

问题原因 #

从kubectl describe service的输出可见Endpoints: ，说明Service未匹配到任何Pod，同时存在两处关键错误：

1. 标签不匹配：Service的selector为name=ems，但ems Pod的labels是app=ems，标签不一致导致Service无法定位目标Pod。
2. 端口不匹配：Service的targetPort设为80，但ems Pod的containerPort为20000，且实际gunicorn API监听在20000端口，端口映射错误会导致流量无法正确转发。

修复方案 #

修改ems-service的YAML配置，修正标签选择器和目标端口：

apiVersion: v1
kind: Service
metadata:
  name: ems-service
spec:
  selector:
    app: ems  # 与Pod的labels保持一致
  ports:
    - protocol: TCP
      port: 20000
      targetPort: 20000  # 与Pod的containerPort保持一致

修改后执行以下命令更新Service：

kubectl apply -f <修改后的service yaml文件路径>

等待Service重新匹配Pod后，即可通过ems-service.default.svc.cluster.local:20000正常访问API。

内容的提问来源于stack exchange，提问作者Divyank