Question

下面的Kubernetes yaml将按预期方式创建资源，并且没有错误/重新启动。到localhost:80005的kafka经纪人的外部请求会无限期挂起。我的理解是，“网络”的码头工人概念是通过服务在Kubernetes中实现的。在这种情况下，我有一个NodePort服务，它将流量引入群集，然后有一个ClusterIp端口，用于处理容器之间的通信。基于此Zookeeper / Kafka配置的非Kubernetes docker run版本，我怀疑我的问题出在我的服务和/或公告的侦听器中。互联网上到处都是docker组成文件和头盔图表，但我似乎无法找到有效的配置组合。

Kafka / Zookeeper YAML：

# Zookeeper
apiVersion: apps/v1
kind: Deployment
metadata:
  name: zookeeper-deploy
spec:
  replicas: 1
  selector:
    matchLabels:
      app: zookeeper-app
  template:
    metadata:
      labels:
        app: zookeeper-app
    spec:
      containers:
        - name: zookeeper-app
          image: confluentinc/cp-zookeeper:5.4.1
          ports:
            - containerPort: 2181
          env:
          - name: ZOOKEEPER_TICK_TIME
            value: "2000"
          - name: ZOOKEEPER_CLIENT_PORT
            value: "2181"
---
apiVersion: v1
kind: Service
metadata:
  name: zk-cluster-svc
  labels:
    app: zk
spec:
  type: ClusterIP
  selector:
    app: zookeeper-app
  ports:
  - port: 2181
    targetPort: 2181
    protocol: TCP
---
# Kafka
apiVersion: apps/v1
kind: Deployment
metadata:
  name: kafka-deploy
spec:
  replicas: 1
  selector:
    matchLabels:
      app: kafka-app
  template:
    metadata:
      labels:
        app: kafka-app
    spec:
      containers:
        - name: kafka-app
          image: confluentinc/cp-kafka:5.4.1
          ports:
            - containerPort: 9092
            - containerPort: 29092
          env:
          - name: KAFKA_BROKER_ID
            value: "1"
          - name: KAFKA_ZOOKEEPER_CONNECT
            value: "zk-cluster-svc:2181"
          - name: KAFKA_ADVERTISED_LISTENERS
            value: "INSIDE://kafka-cluster-svc:29092,OUTSIDE://localhost:30005"
          - name: KAFKA_LISTENER_SECURITY_PROTOCOL_MAP
            value: "INSIDE:PLAINTEXT,OUTSIDE:PLAINTEXT"
          - name: KAFKA_INTER_BROKER_LISTENER_NAME
            value: "INSIDE"
          - name: KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR
            value: "1"
---
apiVersion: v1
kind: Service
metadata:
  name: kafka-service
  labels:
    app: kafka-service
spec:
  type: NodePort
  ports:
  - port: 9092
    nodePort: 30005
    protocol: TCP
  selector:
    app: kafka-app
---
apiVersion: v1
kind: Service
metadata:
  name: kafka-cluster-svc
  labels:
    app: kafka-app
spec:
  selector:
    app: kafka-app
  ports:
  - port: 29092
    targetPort: 29092
    protocol: TCP

^这是基于这些容器的有效docker run版本：

docker network create kafzoonet

docker run --name app-zookeeper -d --network="kafzoonet" --env ZOOKEEPER_TICK_TIME=2000 --env ZOOKEEPER_CLIENT_PORT=2181 confluentinc/cp-zookeeper:5.4.1

docker run --name app-kafka -p 30005:9092 -d --network="kafzoonet" --env KAFKA_BROKER_ID=1 --env KAFKA_ZOOKEEPER_CONNECT="app-zookeeper:2181" --env KAFKA_ADVERTISED_LISTENERS="PLAINTEXT://app-kafka:29092,PLAINTEXT_HOST://localhost:30005" --env KAFKA_LISTENER_SECURITY_PROTOCOL_MAP="PLAINTEXT:PLAINTEXT,PLAINTEXT_HOST:PLAINTEXT" --env KAFKA_INTER_BROKER_LISTENER_NAME=PLAINTEXT --env KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR=1 confluentinc/cp-kafka:5.4.1

是否为单个Kafka节点正确配置了服务？看来Kafka可以到达Zookeeper，但由于某种原因，当客户端连接时，它并没有得到正确的监听器回显。不幸的是，当客户端挂起时，我没有任何错误消息支持此操作，并且Kafka或Zookeeper吊舱都未记录任何内容。

将Kafka docker run命令转换为Kubernetes

0 个答案: