Hadoop集群之Hive HA 安装配置

Hive是基于Hadoop构建的一套数据仓库分析系统，它提供了丰富的SQL查询方式来分析存储在Hadoop 分布式文件系统中的数据。其在Hadoop的架构体系中承担了一个SQL解析的过程，它提供了对外的入口来获取用户的指令然后对指令进行分析，解析出一个 MapReduce程序组成可执行计划，并按照该计划生成对应的MapReduce任务提交给Hadoop集群处理，获取最终的结果。元数据——如表模式 ——存储在名为metastore的数据库中。

1. 安装mysql

①　卸载系统自带的mysql相关安装包，仅卸载mysql开头的包

[hadoop@masternode1 hadoop]# rpm -qa|grep mysql

mysql-libs-5.1.73-3.el6_5.x86_64

[hadoop@masternode1 hadoop]# rpm

rpm            rpmdb          rpmquery       rpmverify      

rpm2cpio       rpmdumpheader  rpmsign        

[hadoop@masternode1 hadoop]# rpm -e --nodeps mysql-libs-5.1.73-3.el6_5.x86_64

②　安装依赖包

[hadoop@masternode1 hadoop]#yum install gcc gcc-c++ ncurses-devel  -y

③　安装cmake

下载cmake安装包

[hadoop@masternode1 cmake-2.8.12]$  wget http://www.cmake.org/files/v2.8/cmake-2.8.12.tar.gz

[hadoop@masternode1 cmake-2.8.12]$  tar zxvf cmake-2.8.12.tar.gz

[hadoop@masternode1 cmake-2.8.12]$ cd cmake-2.8.12

[hadoop@masternode1 cmake-2.8.12]$  ./bootstrap 

[hadoop@masternode1 hadoop]# make && make install

④　创建mysql用户的相应目录

[hadoop@masternode1 ~]# groupadd mysql;useradd -g mysql mysql

[hadoop@masternode1 ~]# su - hadoop

[hadoop@masternode1 cmake-2.8.12]$ mkdir -p /opt/mysql

[hadoop@masternode1 cmake-2.8.12]$ mkdir -p /opt/mysql/data

[hadoop@masternode1 cmake-2.8.12]$ mkdir -p /opt/mysql/log

⑤　修改目录权限

[hadoop@masternode1 ~]$ chmod  +w  /opt/hadoop/mysql-5.6.25

[hadoop@masternode1 ~]$ chmod 756 /opt/hadoop/mysql-5.6.25/ 

⑥　获取MySQL安装包并安装

wget http://dev.mysql.com/get/downloads/mysql/mysql-5.6.25.tar.gz

[hadoop@masternode1 hadoop]]$ tar zxvf mysql-5.6.25.tar.gz

[hadoop@masternode1 hadoop]]$ cd mysql-5.6.25

[hadoop@masternode1 hadoop]]$ mkdir data

[hadoop@masternode1 mysql-5.6.25]$ cmake -DCMAKE_INSTALL_PREFIX=/opt/hadoop/mysql-5.6.25 -DMYSQL_UNIX_ADDR=/opt/hadoop/mysql-5.6.25/mysql.sock -DDEFAULT_CHARSET=utf8 -DDEFAULT_COLLATION=utf8_general_ci -DWITH_INNOBASE_STORAGE_ENGINE=1 -DWITH_ARCHIVE_STORAGE_ENGINE=1 -DWITH_BLACKHOLE_STORAGE_ENGINE=1 -DMYSQL_DATADIR=/opt/hadoop/mysql-5.6.25/data/ -DMYSQL_TCP_PORT=3306 -DENABLE_DOWNLOADS=1-DENABLE_DOWNLOADS=1

若提示cmake命名不能运行，之后执行如下即可

 export PATH=/opt/hadoop/cmake-2.8.12/bin/:$PATH

 [hadoop@masternode1e mysql-5.6.25]# make && make install

注意

DMYSQL_UNIX_ADDR=/opt/hadoop/mysql-5.6.25/mysql.sock  必须目录有写的权限，chmod 756 /opt/hadoop/mysql-5.6.25/ 

或者-DMYSQL_UNIX_ADDR=/tmp/mysql.sock  只需/tmp有写的权限即可 chmod 756 /tmp

Chown mysql:mysql /tmp 因此DMYSQL_UNIX_ADDR最好配置/tmp/mysql.scok

 

⑦　创建软连接

[hadoop@masternode1 ~]#  ln -s /opt/hadoop/mysql-5.6.25/libmysql/libmysqlclient.so.18 /usr/lib/libmysqlclient.so.18

[hadoop@masternode1 ~]# chown hadoop:hadoop  /usr/lib/libmysqlclient.so.18

⑧　修改配置文件

[hadoop@masternode1 mysql-5.6.25]# cp support-files/my-default.cnf /etc/my.cnf

[hadoop@masternode1 mysql-5.6.25]# chown hadoop:hadoop /etc/my.cnf 

[hadoop@masternode1 mysql-5.6.25]$ vi /etc/my.cnf

datadir=/opt/hadoop/mysql-5.6.25/data

log-error=/opt/hadoop/mysql-5.6.25/mysql_error.log

pid-file=/opt/hadoop/mysql-5.6.25/data/mysql.pid

socket=/opt/hadoop/mysql-5.6.25/mysql.sock

user=mysql

⑨　初始化数据库

[hadoop@masternode1 ~]$ chmod 756 /opt/hadoop/mysql-5.6.25/scripts/mysql_install_db

[[[email protected]]$/opt/hadoop/mysql-5.6.25/scripts/mysql_install_db --user=mysql --basedir=/opt/hadoop/mysql-5.6.25 --datadir=/opt/hadoop/mysql-5.6.25/data/  --defaults-file=/etc/my.cnf --basedir=/opt/hadoop/mysql-5.6.25

⑩　创建启动文件

[hadoop@masternode1 ~]# cp /opt/hadoop/mysql-5.6.25/support-files/mysql.server /etc/init.d/mysqld

[hadoop@masternode1 ~]# chown hadoop:hadoop /etc/init.d/mysqld ;chmod 756 /etc/init.d/mysqld

[hadoop@slavenode1 mysql-5.6.25]# mysqld -uroot -h127.0.0.1 -p

bash: mysqld: command not found

解决方法如下：

ln -s /opt/hadoop/mysql-5.6.25/bin/mysql /usr/bin/mysql

 

11　启动MySQL服务

[hadoop@masternode1 mysql-5.6.25]# service mysqld start

Starting MySQL. SUCCESS! 

chkconfig --level 2345 mysqld on

 

12　初始化密码

[hadoop@masternode1 ~]$ mysql -uroot -h127.0.0.1 -p

Enter password: 

Welcome to the MySQL monitor.  Commands end with ; or \g.

Your MySQL connection id is 4

Server version: 5.6.25 Source distribution

 

Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.

 

Oracle is a registered trademark of Oracle Corporation and/or its

affiliates. Other names may be trademarks of their respective

owners.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql> set PASSWORD=PASSWORD('123456');

Query OK, 0 rows affected (0.00 sec)

13　创建Hive用户

mysql> CREATE USER 'hive' IDENTIFIED BY 'hive';

Query OK, 0 rows affected (0.00 sec)

 

mysql> GRANT ALL PRIVILEGES on *.* TO 'hive'@'masternode1' WITH GRANT OPTION;

Query OK, 0 rows affected (0.00 sec)

mysql> flush privileges;

14　Hive用户登录

 [hadoop@masternode1 ~]$ mysql -h masternode1 -uhive

mysql> set PASSWORD = PASSWORD('hive');

15　创建Hive数据库

mysql> create database hive;

Query OK, 1 row affected (0.00 sec)

2. 解压hive软件包

[hadoop@masternode1 hadoop]# tar -zxf apache-hive-2.0.0-bin.tar.gz 

[hadoop@masternode1 hadoop]# mv apache-hive-2.0.0-bin hive-2.0.0

①　Hive环境变量设置

[hadoop@masternode1 hadoop]# cat /etc/profile|tail

#set hive

export HIVE_HOME=/opt/hadoop/hive-2.0.0

export PATH=$PATH:$HIVE_HOME/bin

 vi ~/.bash_profile 

#set hive

export HIVE_HOME=/opt/hadoop/hive-2.0.0

export PATH=$PATH:$HIVE_HOME/bin

②　配置参数文件

[hadoop@masternode1  hadoop$cd hive-2.0.0/conf/

[hadoop@masternode1 conf]$ cp hive-env.sh.template hive-env.sh

[hadoop@masternode1 conf]$ vi hive-env.sh

export HADOOP_HOME=/opt/hadoop/hadoop-2.7.2

export HIVE_HOME=/opt/hadoop/hive-2.0.0

[hadoop@masternode1 conf]$ cp hive-exec-log4j2.properties.template hive-exec-log4j2.properties

[hadoop@masternode1 conf]$ cp hive-log4j2.properties.template hive-log4j2.properties

 

[hadoop@masternode1 conf]$ cp hive-default.xml.template hive-site.xml

[hadoop@masternode1 conf]#  vi hive-site.xml

  hive.metastore.warehouse.dir

  hdfs://cluster-ha/hive/warehouse

  location of default database for the warehouse

  hive.exec.scratchdir

  hdfs://cluster-ha/hive/scratchdir

  Scratch space for Hive jobs

  hive.querylog.location

  /opt/hadoop/hive-2.0.0/logs

  

    Location of Hive run time structured log file

  

  javax.jdo.option.ConnectionURL

  jdbc:mysql://masternode1:3306/hive?createDatabaseIfNotExist=true

  JDBC connect string for a JDBC metastore

  javax.jdo.option.ConnectionDriverName

  com.mysql.jdbc.Driver

  Driver class name for a JDBC metastore

        javax.jdo.option.ConnectionUserName

        

        hive

        username to use against metastore database

    

    

        javax.jdo.option.ConnectionPassword

        

        hive

      password to use against metastore database

     

   hive.exec.local.scratchdir

   /opt/hadoop/hive-2.0.0/local/hive

   Local scratch space for Hive jobs

  

   hive.downloaded.resources.dir

   /opt/hadoop/hive-2.0.0/local/${hive.session.id}_resources

   Temporary local directory for added resources in theremote file system.

  

   hive.server2.logging.operation.log.location

   /opt/hadoop/hive-2.0.0/local/logs/operation_logs

   Top leveldirectory where operation logs are stored if logging functionality isenabled

  

 
  hive.hwi.war.file
  lib/hive-hwi-2.0.1.war
 
 
   hive.hwi.listen.host
     192.168.237.231
       This is the host address the Hive Web Interface will listen on
      
      
         hive.hwi.listen.port
             9999
                 This is the port the Hive Web Interface will listen on
                

  

 

cp hive-exec-log4j.properties.template hive-exec-log4j.properties

vi hive-exec-log4j.properties

hive.log.dir=/opt/hadoop/hive-2.0.0/logs

log4j.appender.EventCounter=org.apache.hadoop.log.metrics.EventCounter

 

cp hive-log4j.properties.template hive-log4j.properties

vi hive-log4j.properties

hive.log.dir=/opt/hadoop/hive-2.0.0/logs

log4j.appender.EventCounter=org.apache.hadoop.log.metrics.EventCounter

③　创建临时目录

[hadoop@masternode1 conf]$ mkdir /opt/hadoop/hive-2.0.0/local/hive

[hadoop@masternode1 hive-2.0.0]$ mkdir local

[hadoop@masternode1 hive-2.0.0]$mkdir /opt/hadoop/hive-2.0.0/logs

[hadoop@masternode1 ~$mkdir /opt/hadoop/hive-2.0.0/local/logs/operation_logs

④　拷贝MySQL驱动文件 

下载地址：

wget http://ftp.nchu.edu.tw/Unix/Database/MySQL/Downloads/Connector-J/mysql-connector-java-5.1.36.tar.gz 

[hadoop@masternode1 conf]# cp /home/centos/mysql-connector-java-5.1.36.tar.gz /opt/hadoop/

[hadoop@masternode1 conf]# cd /opt/hadoop/

[hadoop@masternode1 hadoop]# tar zxvf  mysql-connector-java-5.1.36.tar.gz 

[hadoop@masternode1  hadoop]# cd mysql-connector-java-5.1.36

[hadoop@masternode1 mysql-connector-java-5.1.36]# ls

build.xml  CHANGES  COPYING  docs  mysql-connector-java-5.1.36-bin.jar  README  README.txt  src

[hadoop@masternode1 mysql-connector-java-5.1.36]# cp mysql-connector-java-5.1.36-bin.jar $HIVE_HOME/lib/ 

[hadoop@masternode1 mysql-connector-java-5.1.36]# cd $HIVE_HOME/conf

⑤　拷贝到其他机器上

[hadoop@masternode1 hadoop]# for i in {31,32,33,34,35,36};do scp -r hive-2.0.0/ [email protected]$i:/opt/hadoop/ ; done

⑥　配置７台客户端hive-site.xml

[hadoop@slavenode2 conf]# cat hive-site.xml 

    

  

        hive.metastore.uris  

    thrift://masternode1:9083,thrift://masternode2:9083  

    

⑦　初始化 db 

[hadoop@masternode1 bin]# schematool -initSchema -dbType mysql

Metastore connection URL:        jdbc:mysql://slavenode1:3306/hive_local_meta?createDatabaseIfNotExist=true

Metastore Connection Driver :    com.mysql.jdbc.Driver

Metastore connection User:       hive

Starting metastore schema initialization to 2.0.0

Initialization script hive-schema-2.0.0.mysql.sql

Initialization script completed

schemaTool completed

检查防火墙，不需要得话就关了防火墙，需要得话就把端口放开，比如9083,9000,9001，50070等然后测试，切换到hive主目录，输入一下命令 

[hadoop@masternode1 hive-2.0.0]# bin/hive --service metastore &

Starting Hive Metastore Server

 

⑧　启动hive

$HIVE_HOME/bin/hive

[hadoop@masternode1 hadoop-2.7.2]# $HIVE_HOME/bin/hive

SLF4J: Class path contains multiple SLF4J bindings.

file:/opt/hadoop/hive-2.0.0/conf/hive-log4j2.properties

Hive-on-MR is deprecated in Hive 2 and may not be available in the future versions. Consider using a different execution engine (i.e. tez, spark) or using Hive 1.X releases.

hive> show databases;

OK

hive>  create table studentinfo (id int,name string, age int,tel string) 

    >  row format delimited  fields terminated by '\t' 

    > stored as textfile;

OK

Time taken: 0.439 seconds

hive> load data local inpath '/tmp/stdentInfo.txt'  into table studentinfo;  

Loading data to table default.studentinfo

OK

Time taken: 1.313 seconds

hive> select * from studentinfo; 

OK

1       a       26      110

2       b       29      120 

hive> show tables;

OK

studentinfo

Time taken: 1.008 seconds, Fetched: 1 row(s)

hive> create table tb(id int,name string);

OK

Time taken: 0.427 seconds

 

hadoop@slavenode1 bin]# hive --service hwi

[hadoop@

slavenode1

~]$  /opt/hadoop/hive-2.0.0/bin/hive  --service hwi
hive --service hwils: cannot access /opt/hadoop/hive-2.0.0/lib/hive-hwi-*.war: No such file or directory
SLF4J: Class path contains multiple SLF4J bindings.
SLF4J: Found binding in [jar:file:/opt/hadoop/hive-2.0.0/lib/hive-jdbc-2.0.0-standalone.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/hadoop/hive-2.0.0/lib/log4j-slf4j-impl-2.4.1.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/hadoop/spark/lib/spark-assembly-1.6.1-hadoop2.6.0.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/hadoop/hadoop-2.7.2/share/hadoop/common/lib/slf4j-log4j12-1.7.10.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.
SLF4J: Actual binding is of type [org.apache.logging.slf4j.Log4jLoggerFactory]

解决方法手动构建war包

注意：报错误的原因是，Hive目前的发行包里没有hwi的war包文件，没有这个war包就不能启动hwi服务。

        解决办法：从源码手动build出这个war包，具体过程如第二步。

wget http://mirror.bit.edu.cn/apache/hive/hive-2.0.0/apache-hive-2.0.0-src.tar.gz 

[hadoop@masternode1 hive-2.0.0]$ tar xf apache-hive-2.0.0-src.tar.gz 

[hadoop@masternode1 hive-2.0.0]$ cd apache-hive-2.0.0-src

构建war包

[hadoop@masternode1 hwi]$  jar cvfM0 hive-hwi-2.0.0.war -C web/ . 

 

[hadoop@masternode1 hive-2.0.0]$ cp hive-hwi-2.0.0.war /opt/hadoop/hive-2.0.0/lib/hive-hwi-2.0.0.war

[hadoop@masternode1 hive-2.0.0]$ scp /opt/hadoop/hive-2.0.0/lib/hive-hwi-2.0.0.war hadoop@masternode2:/opt/hadoop/hive-2.0.0/lib/hive-hwi-2.0.0.war

 

hadoop@slavenode1 bin]# hive --service hwi

如有下报错：
Problem accessing /hwi/. Reason:


    Unable to find a javac compiler;
com.sun.tools.javac.Main is not on the classpath.
Perhaps JAVA_HOME does not point to the JDK.
It is currently set to "/usr/java/jdk1.7.0_65/jre"


解决办法：
 cp /usr/java/jdk1.7.0_65/lib/tools.jar  /opt/hadoop/hive-2.0.0/lib/
 hive --service hwi   重启即可。

http://192.168.231.230:9999/hwi/