helm镜像库配置
helm repo add stable http://mirror.azure.cn/kubernetes/charts
helm repo add incubator http://mirror.azure.cn/kubernetes/charts-incubator
helm repo list
NAME URL
stable http://mirror.azure.cn/kubernetes/charts
local http://127.0.0.1:8879/charts
incubator http://mirror.azure.cn/kubernetes/charts-incubator
创建Kafka和Zookeeper的Local PV
创建Kafka的Local PV
这里的部署环境是本地的测试环境，存储选择Local Persistence Volumes。首先，在k8s集群上创建本地存储的StorageClass local-storage.yaml：
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: local-storage
provisioner: kubernetes.io/no-provisioner
volumeBindingMode: WaitForFirstConsumer
reclaimPolicy: Retain
kubectl apply -f local-storage.yaml
storageclass.storage.k8s.io/local-storage created
[root@master home]# kubectl get sc --all-namespaces -o wide
NAME PROVISIONER AGE
local-storage kubernetes.io/no-provisioner 9s
这里要在master,slaver1,slaver2这三个k8s节点上部署3个kafka的broker节点，因此先在三个节点上创建这3个kafka broker节点的Local PV
kafka-local-pv.yaml:
apiVersion: v1
kind: PersistentVolume
metadata:
name: data-kafka-0
spec:
capacity:
storage: 5Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage
local:
path: /home/kafka/data-0
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- master
---
apiVersion: v1
kind: PersistentVolume
metadata:
name: data-kafka-1
spec:
capacity:
storage: 5Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage
local:
path: /home/kafka/data-1
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- slaver1
---
apiVersion: v1
kind: PersistentVolume
metadata:
name: data-kafka-2
spec:
capacity:
storage: 5Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage
local:
path: /home/kafka/data-2
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- slaver2
kubectl apply -f kafka-local-pv.yaml
根据上面创建的local pv，
在master上创建目录/home/kafka/data-0
在slaver1上创建目录/home/kafka/data-1
在slaver2上创建目录/home/kafka/data-2
# master
mkdir -p /home/kafka/data-0
# slaver1
mkdir -p /home/kafka/data-1
# slaver2
mkdir -p /home/kafka/data-2
查看：
[root@master home]# kubectl get pv,pvc --all-namespaces
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
persistentvolume/datadir-kafka-0 5Gi RWO Retain Available local-storage 12s
persistentvolume/datadir-kafka-1 5Gi RWO Retain Available local-storage 12s
persistentvolume/datadir-kafka-2 5Gi RWO Retain Available local-storage 12s
创建Zookeeper的Local PV
这里要在master,slaver1,slaver2这三个k8s节点上部署3个zookeeper节点，因此先在三个节点上创建这3个zookeeper节点的Local PV
zookeeper-local-pv.yaml
apiVersion: v1
kind: PersistentVolume
metadata:
name: data-kafka-zookeeper-0
spec:
capacity:
storage: 5Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage
local:
path: /home/kafka/zkdata-0
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- master
---
apiVersion: v1
kind: PersistentVolume
metadata:
name: data-kafka-zookeeper-1
spec:
capacity:
storage: 5Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage
local:
path: /home/kafka/zkdata-1
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- slaver1
---
apiVersion: v1
kind: PersistentVolume
metadata:
name: data-kafka-zookeeper-2
spec:
capacity:
storage: 5Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage
local:
path: /home/kafka/zkdata-2
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- slaver2
kubectl apply -f zookeeper-local-pv.yaml
根据上面创建的local pv，
在master上创建目录/home/kafka/zkdata-0
在slaver1上创建目录/home/kafka/zkdata-1
在slaver2上创建目录/home/kafka/zkdata-2
# master
mkdir -p /home/kafka/zkdata-0
# slaver1
mkdir -p /home/kafka/zkdata-1
# slaver2
mkdir -p /home/kafka/zkdata-2
查看：
[root@master home]# kubectl get pv,pvc --all-namespaces
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
persistentvolume/data-kafka-zookeeper-0 5Gi RWO Retain Available local-storage 5s
persistentvolume/data-kafka-zookeeper-1 5Gi RWO Retain Available local-storage 5s
persistentvolume/data-kafka-zookeeper-2 5Gi RWO Retain Available local-storage 5s
persistentvolume/datadir-kafka-0 5Gi RWO Retain Available local-storage 116s
persistentvolume/datadir-kafka-1 5Gi RWO Retain Available local-storage 116s
persistentvolume/datadir-kafka-2 5Gi RWO Retain Available local-storage 116s
部署Kafka
编写kafka chart的vaule文件
kafka-values.yaml
replicas: 3
tolerations:
- key: node-role.kubernetes.io/master
operator: Exists
effect: NoSchedule
- key: node-role.kubernetes.io/master
operator: Exists
effect: PreferNoSchedule
persistence:
storageClass: local-storage
size: 5Gi
zookeeper:
persistence:
enabled: true
storageClass: local-storage
size: 5Gi
replicaCount: 3
tolerations:
- key: node-role.kubernetes.io/master
operator: Exists
effect: NoSchedule
- key: node-role.kubernetes.io/master
operator: Exists
effect: PreferNoSchedule
helm install --name kafka --namespace kafka -f kafka-values.yaml incubator/kafka
查看：
[root@master home]# kubectl get po,svc -n kafka -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pod/kafka-0 1/1 Running 2 5m7s 10.244.1.24 slaver1 <none> <none>
pod/kafka-1 1/1 Running 0 2m50s 10.244.2.16 slaver2 <none> <none>
pod/kafka-2 0/1 Running 0 80s 10.244.0.13 master <none> <none>
pod/kafka-zookeeper-0 1/1 Running 0 5m7s 10.244.1.23 slaver1 <none> <none>
pod/kafka-zookeeper-1 1/1 Running 0 4m29s 10.244.2.15 slaver2 <none> <none>
pod/kafka-zookeeper-2 1/1 Running 0 3m43s 10.244.0.12 master <none> <none>
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
service/kafka ClusterIP 10.101.224.127 <none> 9092/TCP 5m7s app=kafka,release=kafka
service/kafka-headless ClusterIP None <none> 9092/TCP 5m7s app=kafka,release=kafka
service/kafka-zookeeper ClusterIP 10.97.247.79 <none> 2181/TCP 5m7s app=zookeeper,release=kafka
service/kafka-zookeeper-headless ClusterIP None <none> 2181/TCP,3888/TCP,2888/TCP 5m7s app=zookeeper,release=kafka
[root@master home]# kubectl get pv,pvc --all-namespaces
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
persistentvolume/data-kafka-zookeeper-0 5Gi RWO Retain Bound kafka/datadir-kafka-2 local-storage 130m
persistentvolume/data-kafka-zookeeper-1 5Gi RWO Retain Bound kafka/data-kafka-zookeeper-0 local-storage 130m
persistentvolume/data-kafka-zookeeper-2 5Gi RWO Retain Bound kafka/data-kafka-zookeeper-1 local-storage 130m
persistentvolume/datadir-kafka-0 5Gi RWO Retain Bound kafka/data-kafka-zookeeper-2 local-storage 132m
persistentvolume/datadir-kafka-1 5Gi RWO Retain Bound kafka/datadir-kafka-0 local-storage 132m
persistentvolume/datadir-kafka-2 5Gi RWO Retain Bound kafka/datadir-kafka-1 local-storage 132m
NAMESPACE NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE
kafka persistentvolumeclaim/data-kafka-zookeeper-0 Bound data-kafka-zookeeper-1 5Gi RWO local-storage 129m
kafka persistentvolumeclaim/data-kafka-zookeeper-1 Bound data-kafka-zookeeper-2 5Gi RWO local-storage 4m36s
kafka persistentvolumeclaim/data-kafka-zookeeper-2 Bound datadir-kafka-0 5Gi RWO local-storage 3m50s
kafka persistentvolumeclaim/datadir-kafka-0 Bound datadir-kafka-1 5Gi RWO local-storage 129m
kafka persistentvolumeclaim/datadir-kafka-1 Bound datadir-kafka-2 5Gi RWO local-storage 2m57s
kafka persistentvolumeclaim/datadir-kafka-2 Bound data-kafka-zookeeper-0 5Gi RWO local-storage 87s
[root@master home]# kubectl get statefulset -n kafka
NAME READY AGE
kafka 3/3 25m
kafka-zookeeper 3/3 25m
安装后的测试
进入一个broker容器查看
[root@master /]# kubectl -n kafka exec kafka-0 -it sh
# ls /usr/bin |grep kafka
kafka-acls
kafka-broker-api-versions
kafka-configs
kafka-console-consumer
kafka-console-producer
kafka-consumer-groups
kafka-consumer-perf-test
kafka-delegation-tokens
kafka-delete-records
kafka-dump-log
kafka-log-dirs
kafka-mirror-maker
kafka-preferred-replica-election
kafka-producer-perf-test
kafka-reassign-partitions
kafka-replica-verification
kafka-run-class
kafka-server-start
kafka-server-stop
kafka-streams-application-reset
kafka-topics
kafka-verifiable-consumer
kafka-verifiable-producer
# ls /usr/share/java/kafka | grep kafka
kafka-clients-2.0.1-cp1.jar
kafka-log4j-appender-2.0.1-cp1.jar
kafka-streams-2.0.1-cp1.jar
kafka-streams-examples-2.0.1-cp1.jar
kafka-streams-scala_2.11-2.0.1-cp1.jar
kafka-streams-test-utils-2.0.1-cp1.jar
kafka-tools-2.0.1-cp1.jar
kafka.jar
kafka_2.11-2.0.1-cp1-javadoc.jar
kafka_2.11-2.0.1-cp1-scaladoc.jar
kafka_2.11-2.0.1-cp1-sources.jar
kafka_2.11-2.0.1-cp1-test-sources.jar
kafka_2.11-2.0.1-cp1-test.jar
kafka_2.11-2.0.1-cp1.jar
可以看到对应apache kafka的版本号是2.11-2.0.1，前面2.11是Scala编译器的版本，Kafka的服务器端代码是使用Scala语言开发的，后边2.0.1是Kafka的版本。 即CP Kafka 5.0.1是基于Apache Kafka 2.0.1的。
安装Kafka Manager
Helm的官方repo中已经提供了Kafka Manager的Chart。
创建kafka-manager-values.yaml：
image:
repository: zenko/kafka-manager
tag: 1.3.3.22
zkHosts: kafka-zookeeper:2181
basicAuth:
enabled: true
username: admin
password: admin
ingress:
enabled: true
hosts:
- km.hongda.com
tls:
- secretName: hongda-com-tls-secret
hosts:
- km.hongda.com
使用helm部署kafka-manager：
helm install --name kafka-manager --namespace kafka -f kafka-manager-values.yaml stable/kafka-manager
安装完成后，确认kafka-manager的Pod已经正常启动：
[root@master home]# kubectl get pod -n kafka -l app=kafka-manager
NAME READY STATUS RESTARTS AGE
kafka-manager-5d974b7844-f4bz2 1/1 Running 0 6m41s
并配置Cluster Zookeeper Hosts为kafka-zookeeper:2181，即可将前面部署的kafka集群纳入kafka-manager管理当中。
查看kafka的po,svc:
[root@master home]# kubectl get po,svc -n kafka -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pod/kafka-0 1/1 Running 4 12d 10.244.1.27 slaver1 <none> <none>
pod/kafka-1 1/1 Running 2 12d 10.244.2.17 slaver2 <none> <none>
pod/kafka-2 1/1 Running 2 12d 10.244.0.15 master <none> <none>
pod/kafka-manager-5d974b7844-f4bz2 1/1 Running 0 43m 10.244.2.21 slaver2 <none> <none>
pod/kafka-zookeeper-0 1/1 Running 1 12d 10.244.1.26 slaver1 <none> <none>
pod/kafka-zookeeper-1 1/1 Running 1 12d 10.244.2.20 slaver2 <none> <none>
pod/kafka-zookeeper-2 1/1 Running 1 12d 10.244.0.16 master <none> <none>
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
service/kafka ClusterIP 10.101.224.127 <none> 9092/TCP 12d app=kafka,release=kafka
service/kafka-headless ClusterIP None <none> 9092/TCP 12d app=kafka,release=kafka
service/kafka-manager ClusterIP 10.102.125.67 <none> 9000/TCP 43m app=kafka-manager,release=kafka-manager
service/kafka-zookeeper ClusterIP 10.97.247.79 <none> 2181/TCP 12d app=zookeeper,release=kafka
service/kafka-zookeeper-headless ClusterIP None <none> 2181/TCP,3888/TCP,2888/TCP 12d app=zookeeper,release=kafka
问题
拉取最新的gcr.io/google_samples/k8szk异常
最新的k8szk版本为3.5.5，拉取不了
解决方法：
docker pull bairuijie/k8szk:3.5.5
docker tag bairuijie/k8szk:3.5.5 gcr.io/google_samples/k8szk:3.5.5
docker rmi bairuijie/k8szk:3.5.5
或者修改镜像版本为k8szk:v3
kubectl edit pod kafka-zookeeper-0 -n kafka
使用k8szk镜像，zookeeper一直处于CrashLoopBackOff状态
[root@master home]# kubectl get po -n kafka -o wide
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
kafka pod/kafka-0 0/1 Running 7 15m 10.244.2.15 slaver2 <none> <none>
kafka pod/kafka-zookeeper-0 0/1 CrashLoopBackOff 7 15m 10.244.1.11 slaver1 <none> <none>
[root@master home]# kubectl logs kafka-zookeeper-0 -n kafka
Error from server: Get https://18.16.202.227:10250/containerLogs/kafka/kafka-zookeeper-0/zookeeper: proxyconnect tcp: net/http: TLS handshake timeout
到slaver1节点上，查看docker容器日志：
[root@slaver1 ~]# docker ps -a
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
51eb5e6e0640 b3a008535ed2 "/bin/bash -xec /con…" About a minute ago Exited (1) About a minute ago k8s_zookeeper_kafka-zookeeper-0_kafka_4448f944-b1cd-4415-8abd-5cee39699b51_8
。。。
[root@slaver1 ~]# docker logs 51eb5e6e0640
+ /config-scripts/run
/config-scripts/run: line 63: /conf/zoo.cfg: No such file or directory
/config-scripts/run: line 68: /conf/log4j.properties: No such file or directory
/config-scripts/run: line 69: /conf/log4j.properties: No such file or directory
/config-scripts/run: line 70: $LOGGER_PROPERS_FILE: ambiguous redirect
/config-scripts/run: line 71: /conf/log4j.properties: No such file or directory
/config-scripts/run: line 81: /conf/log4j.properties: No such file or directory
+ exec java -cp '/apache-zookeeper-*/lib/*:/apache-zookeeper-*/*jar:/conf:' -Xmx1G -Xms1G org.apache.zookeeper.server.quorum.QuorumPeerMain /conf/zoo.cfg
Error: Could not find or load main class org.apache.zookeeper.server.quorum.QuorumPeerMain
查看helm拉取的k8s应用
helm fetch incubator/kafka
ll
-rw-r--r-- 1 root root 30429 8月 23 14:47 kafka-0.17.0.tgz
下载解压以后，发现里面的image依赖的是zookeeper，没有使用k8szk
旧版的kafka-values.yaml:
replicas: 3
persistence:
storageClass: local-storage
size: 5Gi
zookeeper:
persistence:
enabled: true
storageClass: local-storage
size: 5Gi
replicaCount: 3
image:
repository: gcr.io/google_samples/k8szk
去除image绑定即可。
replicas: 3
persistence:
storageClass: local-storage
size: 5Gi
zookeeper:
persistence:
enabled: true
storageClass: local-storage
size: 5Gi
replicaCount: 3
master节点不能部署
查看master节点：
[root@master home]# kubectl describe node master
Name: master
Roles: edge,master
Labels: beta.kubernetes.io/arch=amd64
beta.kubernetes.io/os=linux
kubernetes.io/arch=amd64
kubernetes.io/hostname=master
kubernetes.io/os=linux
node-role.kubernetes.io/edge=
node-role.kubernetes.io/master=
Annotations: flannel.alpha.coreos.com/backend-data: {"VtepMAC":"aa:85:ea:b3:54:07"}
flannel.alpha.coreos.com/backend-type: vxlan
flannel.alpha.coreos.com/kube-subnet-manager: true
flannel.alpha.coreos.com/public-ip: 192.168.236.130
kubeadm.alpha.kubernetes.io/cri-socket: /var/run/dockershim.sock
node.alpha.kubernetes.io/ttl: 0
volumes.kubernetes.io/controller-managed-attach-detach: true
CreationTimestamp: Tue, 20 Aug 2019 23:03:57 +0800
Taints: node-role.kubernetes.io/master:PreferNoSchedule
可以看见master节点被污染了
去除：
kubectl taint nodes master node-role.kubernetes.io/master-
nodes节点后面跟的是节点名称
也可以使用去除所有
kubectl taint nodes --all node-role.kubernetes.io/master-
如果将master设置为train
kubectl taint nodes master node-role.kubernetes.io/master=:NoSchedule
注意⚠️ : 为master设置的这个taint中, node-role.kubernetes.io/master为key, value为空, effect为NoSchedule
如果输入命令时, 你丢掉了=符号, 写成了node-role.kubernetes.io/master:NoSchedule, 会报error: at least one taint update is required错误