HBase 良好的分布式架構(gòu)設(shè)計為海量數(shù)據(jù)的快速存儲、隨機訪問提供了可能，基于數(shù)據(jù)副本機制和分區(qū)機制可以輕松實現(xiàn)在線擴容、縮容和數(shù)據(jù)容災(zāi)，是大數(shù)據(jù)領(lǐng)域中 Key-Value 數(shù)據(jù)結(jié)構(gòu)存儲最常用的數(shù)據(jù)庫方案。
一、概述
HBase 是一個面向列式存儲的分布式數(shù)據(jù)庫，其設(shè)計思想來源于 Google 的 BigTable 論文。HBase 底層存儲基于 HDFS 實現(xiàn)，集群的管理基于 ZooKeeper 實現(xiàn)。HBase 良好的分布式架構(gòu)設(shè)計為海量數(shù)據(jù)的快速存儲、隨機訪問提供了可能，基于數(shù)據(jù)副本機制和分區(qū)機制可以輕松實現(xiàn)在線擴容、縮容和數(shù)據(jù)容災(zāi)，是大數(shù)據(jù)領(lǐng)域中 Key-Value 數(shù)據(jù)結(jié)構(gòu)存儲最常用的數(shù)據(jù)庫方案。
- 官方文檔：https://hbase.apache.org/book.html
- GitHub地址：https://github.com/apache/hbase
- 關(guān)于更多hbase的介紹，也可以參考我這篇文章：列式存儲的分布式數(shù)據(jù)庫——HBase（環(huán)境部署），
- 因為hbase依賴于HDFS存儲，關(guān)于HDFS容器化部署可以參考我以下文章：【云原生】Hadoop on k8s 環(huán)境部署
- 但是生產(chǎn)環(huán)境一般使用高可用的Hadoop的，關(guān)于高可用部署可參考我這篇文章：【云原生】Hadoop HA on k8s 環(huán)境部署
- hbase依賴于Zookeeper，zookeeper容器化部署可以參考：【云原生】zookeeper + kafka on k8s 環(huán)境部署
二、開始編排部署（非高可用HDFS）
地址：https://artifacthub.io/packages/helm/hbase/hbase
1）下載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）構(gòu)建鏡像
在下面連接hadoop高可用會重新構(gòu)建鏡像，這里就不重新構(gòu)建鏡像了，只是把遠程的包推送到本地harbor倉庫
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編排（非高可用HDFS）
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
# 禁用內(nèi)部的hadoop
hadoop:
enabled: false
# 禁用內(nèi)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）開始部署
# 先檢查語法
helm lint ./hbase
# 開始安裝
helm install hbase ./hbase -n hbase --create-namespace
NOTES
NAME: hbase
LAST DEPLOYED: Sat Nov 5 15:44:14 2022
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）測試驗證
測試主備切換，重啟當前active master pod
kubectl delete pod hbase-hbase-master-0 -n hbase
主備能正常切換
6）卸載
helm uninstall hbase -n hbase
# delete ns
kubectl delete ns hbase --force
三、開始編排部署（高可用 HDFS）
1）下載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）構(gòu)建鏡像
這里是基于上面的鏡像進行構(gòu)建，只是把hadoop打包到鏡像中，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
開始構(gòu)建
docker build -t myharbor.com/bigdata/hbase-hdfs-ha:2.4.13.2 . --no-cache
### 參數(shù)解釋
# -t：指定鏡像名稱
# . ：當前目錄Dockerfile
# -f：指定Dockerfile路徑
# --no-cache：不緩存
# 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# 先檢查語法
helm lint ./hbase-hdfs-ha
# 開始安裝
helm install hbase-hdfs-ha ./hbase-hdfs-ha -n hbase-hdfs-ha --create-namespace
NOTES
NAME: hbase-hdfs-ha
LAST DEPLOYED: Sat Nov 5 17:23:20 2022
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）測試驗證
測試主備切換，重啟當前active master pod
kubectl delete pod hbase-hbase-master-0 -n hbase
主備能正常切換
6）卸載
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
