马士兵hadoop2.7.3_hive入门

• Hive入门
• 解压Hive，到/usr/local目录，将解压后的目录名mv为hive 
  设定环境变量HADOOP_HOME，HIVE_HOME，将bin目录加入到PATH中
• 1. cd /usr/local/hive/conf
  2. cp hive-default.xml.template hive-site.xml
  3. 修改hive.metastore.schema.verification，设定为false
  4. 创建/usr/local/hive/tmp目录，替换${system:java.io.tmpdir}为该目录
  5. 替换${system:user.name}为root
• schematool -initSchema -dbType derby 
  会在当前目录下简历metastore_db的数据库。
  注意！！！下次执行hive时应该还在同一目录，默认到当前目录下寻找metastore。
  遇到问题，把metastore_db删掉，重新执行命令
  实际工作环境中，经常使用mysql作为metastore的数据
• 启动hive
• 观察hadoop fs -ls /tmp/hive中目录的创建

• show databases;
  use default;
  create table doc(line string);
  show tables;
  desc doc;
  select * from doc;
  drop table doc;

• 观察hadoop fs -ls /user
• 启动yarn

• load data inpath '/wcinput' overwrite into table doc;
  select * from doc;
  select split(line, ' ') from doc;
  select explode(split(line, ' ')) from doc;
  select word, count(1) as count from (select explode(split(line, ' ')) as word from doc) w group by word;
  select word, count(1) as count from (select explode(split(line, ' ')) as word from doc) w group by word order by word;
  create table word_counts as select word, count(1) as count from (select explode(split(line, ' ')) as word from doc) w group by word order by word;
  select * from word_counts;

  
  
  1. dfs -ls /user/hive/...
• 使用sougou搜索日志做实验
• 将日志文件上传的hdfs系统，启动hive

• create table sougou (qtime string, qid string, qword string, url string) row format delimited fields terminated by ',';
  load data inpath '/sougou.dic' into table sougou;
  select count(*) from sougou;
  create table sougou_results as select keyword, count(1) as count from (select qword as keyword from sougou) t group by keyword order by count desc;
  select * from sougou_results limit 10;