杰霸霸博客

主题

HDFS—集群迁移 ​

Hadoop集群迁移用于升级，或者单纯数据拷贝，常见在3~4年后的Hadoop集群

1. Apache和Apache集群间数据拷贝 ​

使用distcp命令实现两个Hadoop集群之间的递归数据复制

sh
[jack@hadoop102 hadoop-3.3.6]$  bin/hadoop distcp hdfs://hadoop102:8020/user/jack/hello.txt hdfs://hadoop105:8020/user/jack/hello.txt

2. Apache和CDH集群间数据拷贝 ​

2.1 启动Apache和CDH集群 ​

2.2 在apache集群里hosts加上CDH Namenode对应域名并分发给各机器 ​

sh
[root@hadoop101 ~]# vim /etc/hosts

Alt text

2.3 因为集群都是HA模式，所以需要在apache集群上配置CDH集群,让distcp能识别出CDH的nameservice ​

xml
[root@hadoop101 hadoop]# vim /opt/module/hadoop-3.3.6/etc/hadoop/hdfs-site.xml 
<!--配置nameservice-->
<property>
  <name>dfs.nameservices</name>
  <value>mycluster,nameservice1</value>
</property>

<!--指定本地服务-->
<property>
  <name>dfs.internal.nameservices</name>
  <value>mycluster,nameservice1</value>
</property>
<!--配置多NamenNode-->
<property>
  <name>dfs.ha.namenodes.mycluster</name>
  <value>nn1,nn2,nn3</value>
</property>
<property>
  <name>dfs.namenode.rpc-address.mycluster.nn1</name>
  <value>hadoop101:8020</value>
</property>
<property>
<property>
  <name>dfs.namenode.rpc-address.mycluster.nn2</name>
  <value>hadoop102:8020</value>
</property>
<property>
  <name>dfs.namenode.rpc-address.mycluster.nn3</name>
  <value>hadoop103:8020</value>
</property>
<!--配置nameservice1的namenode服务-->
<property>
    <name>dfs.ha.namenodes.nameservice1</name>
    <value>namenode30,namenode37</value>
  </property>
 <property>
    <name>dfs.namenode.rpc-address.nameservice1.namenode30</name>
    <value>hadoop104:8020</value>
  </property>
<property>
    <name>dfs.namenode.rpc-address.nameservice1.namenode37</name>
    <value>hadoop106:8020</value>
  </property>
<property>
    <name>dfs.namenode.http-address.nameservice1.namenode30</name>
    <value>hadoop104:9870</value>
  </property>
  <property>
    <name>dfs.namenode.http-address.nameservice1.namenode37</name>
    <value>hadoop106:9870</value>
  </property>
  <property>
    <name>dfs.client.failover.proxy.provider.nameservice1</name>
    <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
  </property>
<!--为NamneNode设置HTTP服务监听-->
<property>
  <name>dfs.namenode.http-address.mycluster.nn1</name>
  <value>hadoop101:9870</value>
</property>
<property>
  <name>dfs.namenode.http-address.mycluster.nn2</name>
  <value>hadoop102:9870</value>
</property>
<property>
  <name>dfs.namenode.http-address.mycluster.nn3</name>
  <value>hadoop103:9870</value>
</property>
<!--配置HDFS客户端联系Active NameNode节点的Java类-->
<property>
  <name>dfs.client.failover.proxy.provider.mycluster</name>
  <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
</property>

2.4 修改CDH hosts ​

sh
[root@hadoop101 ~]# vim /etc/hosts

Alt text 进行分发，这里的hadoop104，hadoop105，hadoop106分别对应apache的hadoop101，hadoop102，hadoop103

2.5 修改CDH集群配置，在所有hdfs-site.xml文件里修改配置 ​

Alt text

xml
<property>
	<name>dfs.nameservices</name>
	<value>mycluster,nameservice1</value>
</property>

<property>
	<name>dfs.internal.nameservices</name>
	<value>nameservice1</value>
</property>

<property>
	<name>dfs.ha.namenodes.mycluster</name>
	<value>nn1,nn2,nn3</value>
</property>

<property>
	<name>dfs.namenode.rpc-address.mycluster.nn1</name>
	<value>hadoop104:8020</value>
</property>

<property>
	<name>dfs.namenode.rpc-address.mycluster.nn2</name>
	<value>hadoop105:8020</value>
</property>

<property>
	<name>dfs.namenode.rpc-address.mycluster.nn3</name>
	<value>hadoop106:8020</value>
</property>

<property>
	<name>dfs.namenode.http-address.mycluster.nn1</name>
	<value>hadoop104:9870</value>
</property>

<property>
	<name>dfs.namenode.http-address.mycluster.nn2</name>
	<value>hadoop105:9870</value>
</property>

<property>
	<name>dfs.namenode.http-address.mycluster.nn3</name>
	<value>hadoop106:9870</value>
</property>

<property>
	<name>dfs.client.failover.proxy.provider.mycluster</name>
	<value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
</property>

注意

重点由于我的Apahce集群和CDH集群3台集群都是hadoop101，hadoop102，hadoop103所以要关闭域名访问，使用IP访问

2.6 CDH把钩去了 ​

Alt text

2.7 apache设置为false ​

Alt text

2.8 再使用hadoop distcp命令进行迁移，-Dmapred.job.queue.name指定队列，默认是default队列。上面配置集群都配了的话，那么在CDH和apache集群下都可以执行这个命令 ​

sh
[root@hadoop101 hadoop]# hadoop distcp -Dmapred.job.queue.name=hive  webhdfs://mycluster:9070/user/hive/warehouse/dwd.db/  hdfs://nameservice1/user/hive/warehouse

Alt text 执行会启动一个MR任务，正在迁移 Alt text 查看CDH 9870 http地址 Alt textAlt text

2.9 数据已经成功迁移。数据迁移成功之后，接下来迁移hive表结构，编写shell脚本 ​

sh
[root@hadoop101 module]# vim exportHive.sh 
#!/bin/bash
hive -e "use dwd;show tables">tables.txt
cat tables.txt |while read eachline
do
hive -e "use dwd;show create table $eachline">>tablesDDL.txt
echo ";" >> tablesDDL.txt
done

2.10 执行脚本后将tablesDDL.txt文件分发到CDH集群下 ​

sh
[root@hadoop101 module]# scp tablesDDL.txt  hadoop104:/opt/module/

2.11 然后CDH下导入此表结构，先进到CDH的hive里创建dwd库 ​

sh
[root@hadoop101 module]# hive
hive> create database dwd;

2.12 创建数据库后，边界tablesDDL.txt在最上方加上use dwd; ​

Alt text

2.13 将createtab_stmt都替换成空格 ​

sh
[root@hadoop101 module]# sed -i s"#createtab_stmt# #g" tablesDDL.txt

2.14 最后执行hive -f命令将表结构导入 ​

sh
[root@hadoop101 module]# hive -f tablesDDL.txt

Alt text

2.15 最后将表的分区重新刷新下，只有刷新分区才能把数据读出来，编写脚本 ​

sh
[root@hadoop101 module]# vim msckPartition.sh
#!/bin/bash
hive -e "use dwd;show tables">tables.txt
cat tables.txt |while read eachline
do
hive -e "use dwd;MSCK REPAIR TABLE $eachline"
done
[root@hadoop101 module]# chmod +777 msckPartition.sh 
[root@hadoop101 module]# ./msckPartition.sh

2.16 刷完分区后，查询表数据 ​

Alt text