一、环境依赖
Hive必须运行在Hadoop之上,则需要先安装Hadoop环境
二、安装Hive
1.下载Hive
我安装的是apache-hive-1.2.1-bin.tar.gz
[hadoop@hftclclw0001 ~]$ pwd/home/hadoop[hadoop@hftclclw0001 ~]$ wget http://apache.mirrors.ionfish.org/hive/hive-1.2.1/apache-hive-1.2.1-bin.tar.gz[hadoop@hftclclw0001 ~]$ lltotal 637256drwx------ 10 hadoop root 4096 Oct 27 02:22 apache-hive-1.2.1-bin-rw------- 1 hadoop root 92834839 Jun 26 18:34 apache-hive-1.2.1-bin.tar.gzdrwx------ 3 hadoop root 4096 Oct 27 09:05 datadrwx------ 11 hadoop root 4096 Oct 21 03:20 hadoop-2.7.1-rw------- 1 hadoop root 210606807 Oct 20 09:00 hadoop-2.7.1.tar.gzdrwx------ 2 hadoop root 4096 Oct 23 02:08 install-sqoopdrwx------ 13 hadoop root 4096 Oct 20 09:22 spark-1.5.1-bin-hadoop2.6-rw------- 1 hadoop root 280901736 Oct 20 09:19 spark-1.5.1-bin-hadoop2.6.tgzdrwx------ 22 hadoop root 4096 Oct 23 02:08 sqoop-1.99.6-bin-hadoop200-rw------- 1 hadoop root 68177818 May 5 22:34 sqoop-1.99.6-bin-hadoop200.tar.gz[hadoop@hftclclw0001 ~]$ cd apache-hive-1.2.1-bin/conf/[hadoop@hftclclw0001 conf]$ pwd/home/hadoop/apache-hive-1.2.1-bin/conf[hadoop@hftclclw0001 conf]$ vi hive-env.sh......HADOOP_HOME=/home/hadoop/hadoop-2.7.1 =>配置Hadoop_Homeexport HIVE_CONF_DIR=/home/hadoop/apache-hive-1.2.1-bin/conf =>配置HIVE_conf_homeexport HIVE_AUX_JARS_PATH=/home/hadoop/apache-hive-1.2.1-bin/lib/ #我使用了mysql作为metastore ,则需要在lib目录下添加mysql的驱动[hadoop@hftclclw0001 lib]$ pwd/home/hadoop/apache-hive-1.2.1-bin/lib[hadoop@hftclclw0001 lib]$ ll | grep mysql-rw------- 1 hadoop root 848401 Oct 27 01:48 mysql-connector-java-5.1.25-bin.jar[hadoop@hftclclw0001 conf]$ vi hive-site.xml[hadoop@hftclclw0001 conf]$ cat hive-site.xml[hadoop@hftclclw0001 conf]$ vi hive-log4j.properties => Log4j的配置,可以修改日志目录... hive.metastore.local false =>metastore我的mysql不是在该server上javax.jdo.option.ConnectionURL jdbc:mysql://{ip:port}/{databases} => mysql服务的ip和端口号javax.jdo.option.ConnectionDriveName com.mysql.jdbc.Driver javax.jdo.option.ConnectionUserName {username} javax.jdo.option.ConnectionPassword {password} hive.metastore.warehouse.dir /hive/warehouse =>hive的仓库目录,需要在HDFS上创建,并修改权限hive.metastore.uris thrift://{ip}:{port} =>本机ip和端口号,启动metastore服务
2.启动metastore
[hadoop@hftclclw0001 bin]$ pwd/home/hadoop/apache-hive-1.2.1-bin/bin[hadoop@hftclclw0001 bin]$ ./hive --service metastore &[hadoop@hftclclw0001 bin]$ ps ax|grep metastore......
3.启动HiveServer2
[hadoop@hftclclw0001 bin]$ pwd/home/hadoop/apache-hive-1.2.1-bin/bin[hadoop@hftclclw0001 bin]$ ./hive --service hiveserver2 &[hadoop@hftclclw0001 bin]$ ps ax|grep HiveServer2......
4.启动shell 或是 beeline
[hadoop@hftclclw0001 bin]$ ./hive shell......
三、metastore
1.内置模式:将数据保存在内置的Derby数据库中,这种方式最简单,但是Derby每次只能访问一个数据文件。
Drive ==> Metastore ==> Derby
2.本地模式:将元数据保存在本地的独立数据库(如mysql)等
Driver ===> Metastore
Driver ===> Metastore ===> DB
Driver ===> Metastore
每个server都需要配置metastore,并启动metastore服务
3.远程模式:使用thrift访问metastore
Client1
Client2 ===> Metastore ===> DB
Client3
4.配置:
如上述配置,我们已经启动了metastore服务在上述hftclclw0001机器上,我们在另一台server上,如hftclcld0001机器上,安装hive,配置如上述配置一直,仅仅修改hive-site.xml如下:即我们通过thrift协议,访问hftclclw0001上面的metastore,并访问hive的元数据[root@hftclcld0001 apache-hive-1.2.1-bin]# pwd/home/hadoop/apache-hive-1.2.1-bin[root@hftclcld0001 apache-hive-1.2.1-bin]# ./bin/hive shellhive> hive> show databases;OKdefaulthive =>能访问到hive的metastore,访问到元数据(我们之前创建的)human_resourcesTime taken: 0.388 seconds, Fetched: 3 row(s)hive> hive.metastore.uris thrift://{ip}:{port} =>hftclclw0001机器的ip和端口号