【树叶云】HBase on K8s 编排部署讲解与实战操作

一、概述

HBase 是一个面向列式存储的分布式数据库，其设计思想来源于 Google 的 BigTable 论文。HBase 底层存储基于 HDFS 实现，集群的管理基于 ZooKeeper 实现。HBase 良好的分布式架构设计为海量数据的快速存储、随机访问提供了可能，基于数据副本机制和分区机制可以轻松实现在线扩容、缩容和数据容灾，是大数据领域中 Key-Value 数据结构存储最常用的数据库方案。

• 官方文档：https://hbase.apache.org/book.html
• GitHub地址：https://github.com/apache/hbase
• 关于更多hbase的介绍，也可以参考我这篇文章：列式存储的分布式数据库——HBase（环境部署），
• 因为hbase依赖于HDFS存储，关于HDFS容器化部署可以参考我以下文章：【云原生】Hadoop on k8s 环境部署
• 但是生产环境一般使用高可用的Hadoop的，关于高可用部署可参考我这篇文章：【云原生】Hadoop HA on k8s 环境部署
• hbase依赖于Zookeeper，zookeeper容器化部署可以参考：【云原生】zookeeper + kafka on k8s 环境部署

二、开始编排部署（非高可用HDFS）

地址：https://artifacthub.io/packages/helm/hbase/hbase

1）下载chart 包

helm repo add hbase https://itboy87.github.io/bigdata-charts/

# hbase version 2.4.13
helm pull hbase/hbase --version 0.1.7

2）构建镜像

在下面连接hadoop高可用会重新构建镜像，这里就不重新构建镜像了，只是把远程的包推送到本地harbor仓库

docker pull ghcr.io/fleeksoft/hbase/hbase-base:2.4.13.2

# tag
docker tag ghcr.io/fleeksoft/hbase/hbase-base:2.4.13.2 myharbor.com/bigdata/hbase-base:2.4.13.2

# push
docker push myharbor.com/bigdata/hbase-base:2.4.13.2

3）修改yaml编排（非高可用HDFS）

• hbase/values.yaml

image:
  repository: myharbor.com/bigdata/hbase-base
  tag:2.4.13.2
  pullPolicy: IfNotPresent

...

conf:
  hadoopUserName: admin
  hbaseSite:
    hbase.rootdir:"hdfs://hadoop-hadoop-hdfs-nn.hadoop:9000/hbase"
    hbase.zookeeper.quorum:"zookeeper.zookeeper:2181"

...

hbase:
  master:
    replicas:2

  regionServer:
    replicas:2

# 禁用内部的hadoop
hadoop:
  enabled:false

# 禁用内部的zookeeper
zookeeper:
  enabled:false

• hbase/templates/hbase-configmap.yaml

if [ ` `.`Values`.`hadoop`.`enabled ` = true ];then
  NAMENODE_URL={{- printf "http://%s-hadoop-hdfs-nn:9870/index.html" .Release.Name }}
else
  hadoop_url={{ index .Values.conf.hbaseSite "hbase.rootdir" }}
  hadoop_url=`echo $hadoop_url|awk -F '/' '{print $3}'|awk -F':' '{print $1}'`
  NAMENODE_URL=http://${hadoop_url}:9870/index.html
fi

4）开始部署

# 先检查语法
helm lint ./hbase

# 开始安装
helm install hbase ./hbase -n hbase --create-namespace

NOTES

NAME: hbase
LAST DEPLOYED: Sat Nov  515:44:142022
NAMESPACE: hbase
STATUS: deployed
REVISION:1
TEST SUITE: None
NOTES:
1. You can get an HBASE Shell by running this command:
   kubectl exec -n hbase -it hbase-hbase-master-0-- hbase shell

2. Inspect hbase master service ports with:
   kubectl exec -n hbase describe service hbase-hbase-master

3. Create a port-forward to the hbase manager UI:
   kubectl port-forward -n hbase svc/hbase-hbase-master 16010:16010

   Then open the ui in your browser:

   open http://localhost:16010

4. Create a port-forward to the hbase thrift manager UI:
   kubectl port-forward -n hbase svc/hbase-hbase-master 9095:9095

   Then open the ui in your browser:

   open http://localhost:9095

HDFS

查看

kubectl get pods,svc -n hbase -owide

5）测试验证

测试主备切换，重启当前active master pod

kubectl delete pod hbase-hbase-master-0-n hbase

主备能正常切换

6）卸载

helm uninstall hbase -n hbase
# delete ns 
kubectl delete ns hbase --force

三、开始编排部署（高可用 HDFS）

1）下载chart 包

helm repo add hbase https://itboy87.github.io/bigdata-charts/

# hbase version 2.4.13
helm pull hbase/hbase --version 0.1.7

2）构建镜像

这里是基于上面的镜像进行构建，只是把hadoop打包到镜像中，主要用的hadoop配置文件是core-site.yaml，hdfs-site.yaml

Dockerfile

FROM myharbor.com/bigdata/hbase-base:2.4.13.2

RUN mkdir -p /opt/apache

ENV HADOOP_VERSION=3.3.2

ADD hadoop-${HADOOP_VERSION}.tar.gz/opt/apache

ENV HADOOP_HOME=/opt/apache/hadoop

RUN ln -s /opt/apache/hadoop-${HADOOP_VERSION} $HADOOP_HOME

ENV HADOOP_CONF_DIR=${HADOOP_HOME}/et/hadoop

ENV PATH=${HADOOP_HOME}/bin:$PATH

开始构建

docker build -t myharbor.com/bigdata/hbase-hdfs-ha:2.4.13.2 . --no-cache

### 参数解释
# -t：指定镜像名称
# . ：当前目录Dockerfile
# -f：指定Dockerfile路径
#  --no-cache：不缓存

# 推送到harbor
docker push myharbor.com/bigdata/hbase-hdfs-ha:2.4.13.2

3）修改配置

• hbase-hdfs-ha/values.yaml

image:
  repository: myharbor.com/bigdata/hbase-hdfs-ha
  tag:2.4.13.2
  pullPolicy: IfNotPresent

...

conf:
  hadoopUserName: admin
  hbaseSite:
    hbase.rootdir:"hdfs://myhdfs/hbase"
    hbase.zookeeper.quorum:"zookeeper.zookeeper:2181"

• hbase-hdfs-ha/templates/hbase-configmap.yaml

if [{{.Values.hadoop.enabled}}=true];then
  NAMENODE_URL={{- printf "http://%s-hadoop-hdfs-nn:9870/index.html".Release.Name}}
else
  NAMENODE_URL=http://hadoop-ha-hadoop-hdfs-nn-1.hadoop-ha:9870:9870/index.html
fi

# 先检查语法
helm lint ./hbase-hdfs-ha

# 开始安装
helm install hbase-hdfs-ha ./hbase-hdfs-ha -n hbase-hdfs-ha --create-namespace

NOTES

NAME: hbase-hdfs-ha
LAST DEPLOYED: Sat Nov  517:23:202022
NAMESPACE: hbase-hdfs-ha
STATUS: deployed
REVISION:1
TEST SUITE: None
NOTES:
1. You can get an HBASE Shell by running this command:
   kubectl exec -n hbase-hdfs-ha -it hbase-hdfs-ha-hbase-master-0-- hbase shell

2. Inspect hbase master service ports with:
   kubectl exec -n hbase-hdfs-ha describe service hbase-hdfs-ha-hbase-master

3. Create a port-forward to the hbase manager UI:
   kubectl port-forward -n hbase-hdfs-ha svc/hbase-hdfs-ha-hbase-master 16010:16010

   Then open the ui in your browser:

   open http://localhost:16010

4. Create a port-forward to the hbase thrift manager UI:
   kubectl port-forward -n hbase-hdfs-ha svc/hbase-hdfs-ha-hbase-master 9095:9095

   Then open the ui in your browser:

   open http://localhost:9095

HDFS

查看

kubectl get pods,svc -n hbase-hdfs-ha

5）测试验证

测试主备切换，重启当前active master pod

kubectl delete pod hbase-hbase-master-0-n hbase

主备能正常切换

6）卸载

helm uninstall hbase-hdfs-ha -n hbase-hdfs-ha
# delete ns 
kubectl delete ns hbase-hdfs-ha --force

git 地址：https://gitee.com/hadoop-bigdata/hbase-on-k8s