Question

我的minikube中的kafka和zookeeper具有以下配置：

apiVersion: v1
kind: Service
metadata:
  name: kafka-service
  namespace: kafka
spec:
  selector:
    app: kafka
  ports:
    - protocol: TCP
      port: 9092
      name: kafka-port
    - protocol: TCP
      port: 9094
      name: kafka-port-out
    - protocol: TCP
      port: 2181
      name: kafka-zk
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: kafka-deployment
  namespace: kafka
spec:
  replicas: 1
  selector:
    matchLabels:
      app: kafka
      k8s-app: kube-dns
  template:
    metadata:
      labels:
        app: kafka
        k8s-app: kube-dns
    spec:
      containers:
        - name: kafka-container
          image: bitnami/kafka:latest
          env:
          - name: 'ALLOW_PLAINTEXT_LISTENER'
            value: 'yes'
          - name: 'KAFKA_CFG_ZOOKEEPER_CONNECT'
            value: 'zookeeper-service:2181'
          - name: 'KAFKA_CFG_LISTENERS'
            value: 'PLAINTEXT://:9092'
          - name: 'KAFKA_CFG_ADVERTISED_LISTENERS' # if I comment this and the next line it works only locally
            value: 'PLAINTEXT://kafka-service.kafka:9092'
          ports:
            - containerPort: 9092
              name: kafka-port
            - containerPort: 9094
              name: kafka-port-out
            - containerPort: 5555
              name: kafka-port-jmx
            - containerPort: 2181
              name: kafka-zk

这是我的动物园管理员的配置：

apiVersion: v1
kind: Service
metadata:
  name: zookeeper-service
  namespace: kafka
spec:
  selector:
    app: zookeeper
  ports:
    - protocol: TCP
      port: 2181
      name: zookeeper-port
    - protocol: TCP
      port: 2888
      name: zookeeper-peer
    - protocol: TCP
      port: 3888
      name: leader-election
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: zookeeper-deployment
  namespace: kafka
spec:
  replicas: 1
  selector:
    matchLabels:
      app: zookeeper
      k8s-app: kube-dns
  template:
    metadata:
      labels:
        app: zookeeper
        k8s-app: kube-dns
    spec:
      containers:
        - name: zookeeper-container
          image: bitnami/zookeeper:latest
          env:
            - name: 'ALLOW_ANONYMOUS_LOGIN'
              value: 'yes'
            - name: 'ZOOKEEPER_ID'
              value: '1'
          ports:
            - containerPort: 2181
              name: zookeeper-port
            - containerPort: 2888
              name: zookeeper-peer
            - containerPort: 3888
              name: leader-election

我还有另一个部署：

apiVersion: apps/v1
kind: Deployment
metadata:
  name: kafkacat-deployment
#  namespace: debug # I was using it from another namespace, but it was not working so I've tried to use the same
  namespace: kafka
spec:
  replicas: 1
  selector:
    matchLabels:
      app: kafkacat
      k8s-app: kube-dns
  template:
    metadata:
      labels:
        app: kafkacat
        k8s-app: kube-dns
    spec:
      containers:
      - name: kafkacat-container
        image: edenhill/kafkacat:1.5.0

然后我尝试通过telnet进行连接，并且有效。

telnet kafka-service.kafka 9092
Trying 10.101.87.127...
Connected to kafka-service.kafka.svc.cluster.local.
Escape character is '^]'.

这是nslookup

nslookup kafka-service.kafka
Server:     10.96.0.10
Address:    10.96.0.10#53

Name:   kafka-service.kafka.svc.cluster.local
Address: 10.101.87.127

但是当我尝试达到它时，我得到的是：

kafkacat -b kafka-service.kafka:9092 -L
% ERROR: Failed to acquire metadata: Local: Timed out

所以我猜想问题出在Kafka配置中：如果我评论env var KAFKA_CFG_ADVERTISED_LISTENERS，它的工作原理如下：

# kafkacat -b kafka-service.kafka:9092 -L
Metadata for all topics (from broker -1: kafka-service.kafka:9092/bootstrap):
 1 brokers:
  broker 1001 at kafka-deployment-858c5c7f98-tt7sr:9092
 2 topics:
  topic "my_topic" with 1 partitions:
    partition 0, leader 1001, replicas: 1001, isrs: 1001
  topic "__consumer_offsets" with 50 partitions:
    partition 0, leader 1001, replicas: 1001, isrs: 1001

如果我尝试产生一条消息：

kafkacat -b kafka-service.kafka:9092 -P -t my_topic
oi
% ERROR: Local: Host resolution failure: kafka-deployment-858c5c7f98-w2dm5:9092/1001: Failed to resolve 'kafka-deployment-858c5c7f98-w2dm5:9092': Temporary failure in name resolution (after 15382734ms in state INIT)

然后，如果我尝试食用：

kafkacat -b kafka-service.kafka:9092 -C -t my_topic
% ERROR: Local: Host resolution failure: kafka-deployment-858c5c7f98-w2dm5:9092/1001: Failed to resolve 'kafka-deployment-858c5c7f98-w2dm5:9092': Temporary failure in name resolution (after 15406287ms in state INIT)

我尝试配置KAFKA_CFG_ADVERTISED_LISTENERS=PLAINTEXT://kafka-service.kafka.svc.cluster.local:9092，但尝试通过kafkacat获取信息时仍然会超时。

这就是为什么播发的侦听器是无法从本地计算机访问的主机名的原因。我该如何解决群集中的Kafka配置？

Answer 1

我应用了与您提供的清单完全相同的清单，除了在kafka-deployment pod上的/ dev / null中添加了一条尾巴，并且能够产生和使用主题。

我正在k3d中运行一个k3s集群：

└[~]> kubectl get nodes                                                    
NAME               STATUS   ROLES    AGE   VERSION
k3d-dev-server     Ready    master   31m   v1.17.3+k3s1
k3d-dev-worker-1   Ready    <none>   31m   v1.17.3+k3s1

kafka部署清单：

apiVersion: apps/v1
kind: Deployment
metadata:
  name: kafkacat-deployment
  #  namespace: debug # I was using it from another namespace, but it was not working so I've tried to use the same
  namespace: kafka
spec:
  replicas: 1
  selector:
    matchLabels:
      app: kafkacat
      k8s-app: kube-dns
  template:
    metadata:
      labels:
        app: kafkacat
        k8s-app: kube-dns
    spec:
      containers:
        - name: kafkacat-container
          image: edenhill/kafkacat:1.5.0
          resources: {}
          command:
            - sh
            - -c
            - "exec tail -f /dev/null"

Kafka命名空间中来自kafka-deployment pod的终端日志：

/ # kafkacat -b kafka-service.kafka:9092 -L
Metadata for all topics (from broker 1001: kafka-service.kafka:9092/1001):
 1 brokers:
  broker 1001 at kafka-service.kafka:9092 (controller)
 0 topics:
/ # kafkacat -b kafka-service.kafka:9092 -P -t my_topic
hi from kafkacat

/ # kafkacat -b kafka-service.kafka:9092 -C -t my_topic
hi from kafkacat
% Reached end of topic my_topic [0] at offset 1

/ # kafkacat -b kafka-service:9092 -L
Metadata for all topics (from broker -1: kafka-service:9092/bootstrap):
 1 brokers:
  broker 1001 at kafka-service.kafka:9092 (controller)
 1 topics:
  topic "my_topic" with 1 partitions:
    partition 0, leader 1001, replicas: 1001, isrs: 1001
/ # kafkacat -b kafka-service:9092 -P -t my_topic
hi from kafka2

/ # kafkacat -b kafka-service:9092 -C -t my_topic
hi from kafkacat
hi from kafka2
% Reached end of topic my_topic [0] at offset 2

来自调试名称空间的kafka部署的终端日志：

└[~]> kubectl exec -it kafkacat-deployment-76f9c9db6d-8fth4 -n debug -- ash
/ # kafkacat -b kafka-service.kafka:9092 -L
Metadata for all topics (from broker 1001: kafka-service.kafka:9092/1001):
 1 brokers:
  broker 1001 at kafka-service.kafka:9092 (controller)
 1 topics:
  topic "my_topic" with 1 partitions:
    partition 0, leader 1001, replicas: 1001, isrs: 1001
/ # kafkacat -b kafka-service.kafka:9092 -P -t my_topic
hi from debug namespace

/ # kafkacat -b kafka-service.kafka:9092 -C -t my_topic
hi from kafkacat
hi from kafka2
hi from debug namespace
% Reached end of topic my_topic [0] at offset 3

在针对minikube中的错误应用了解决方法之后，我也使它也可以在minikube中工作：

minikube ssh
sudo ip link set docker0 promisc on

删除了正在运行的kafka-deployment pod，它可以在两个名称空间中使用：

(⎈ |minikube:default)➜  ~ kubectl delete pod -n kafka kafka-deployment-5c4f64599f-kn9xt
pod "kafka-deployment-5c4f64599f-kn9xt" deleted
(⎈ |minikube:default)➜  ~ kubectl exec -n kafka -it kafkacat-deployment-b595d9ccd-4bht7 -- ash
/ # kafkacat -b kafka-service.kafka:9092 -L
Metadata for all topics (from broker 1003: kafka-service.kafka:9092/1003):
 1 brokers:
  broker 1003 at kafka-service.kafka:9092 (controller)
 0 topics:
/ # 
(⎈ |minikube:default)➜  ~ kubectl exec -n debug -it kafkacat-deployment-b595d9ccd-pgzv6 -- ash
/ # kafkacat -b kafka-service.kafka:9092 -L
Metadata for all topics (from broker 1003: kafka-service.kafka:9092/1003):
 1 brokers:
  broker 1003 at kafka-service.kafka:9092 (controller)
 0 topics:
/ #

从k8s中的另一个命名空间连接到kafka

1 个答案: