在 Hive 0.14 之前，Hive QL 一直不支持insert、update、delete 操作，这显然很不方便，尤其是在构建数据仓库的过程中，一个比较常见的例子是维度表经常需要更新某列，在 Hive 中需要更新历史所有数据，这显然是不合理的。
在 Hive 0.14 版本，支持了行级别的 ACID 与 Transactions，这也就解决了上面的问题。
本文主要讲解如何在 Hive 0.14 配置，使得支持上述两个特性，相关原理及性能分析后面再单独写一篇文章来分析。
相关软件说明：
- Hive 1.1.0
- mysql 5.6（存储 metastore ）
1. 初始化数据库
mysql> create database metastore_db;
mysql> USE metastore_db;
mysql> SOURCE $HIVE_HOME/scripts/metastore/upgrade/mysql/hive-schema-1.1.0.mysql.sql
mysql> SOURCE $HIVE_HOME/scripts/metastore/upgrade/mysql/hive-txn-schema-0.13.0.mysql.sql
mysql> CREATE USER 'hive'@'localhost' IDENTIFIED BY 'mypassword';
mysql> grant all privileges on metastore_db.* to hive@'%' identified by 'hive';
mysql> FLUSH PRIVILEGES;
mysql> quit; 这里需要注意的是，必须source txn相关的表，否则后面创建表是会报下面几种错误：
FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. MetaException(message:javax.jdo.JDODataStoreException: Schema Transaction threw exception "Add classes to Catalog "metastore_db", Schema """
at org.datanucleus.api.jdo.NucleusJDOHelper.getJDOExceptionForNucleusException(NucleusJDOHelper.java:451)
at org.datanucleus.api.jdo.JDOPersistenceManager.jdoMakePersistent(JDOPersistenceManager.java:732)
at org.datanucleus.api.jdo.JDOPersistenceManager.makePersistent(JDOPersistenceManager.java:752)
at org.apache.hadoop.hive.metastore.ObjectStore.createTable(ObjectStore.java:794)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:497)
at org.apache.hadoop.hive.metastore.RawStoreProxy.invoke(RawStoreProxy.java:98)
at com.sun.proxy.$Proxy6.createTable(Unknown Source)
at org.apache.hadoop.hive.metastore.HiveMetaStore$HMSHandler.create_table_core(HiveMetaStore.java:1393)
at org.apache.hadoop.hive.metastore.HiveMetaStore$HMSHandler.create_table_with_environment_context(HiveMetaStore.java:1426)2. 配置 hive-site.xml
<property>
<name>hive.support.concurrency</name>
<value>true</value>
</property>
<property>
<name>hive.txn.manager</name>
<value>org.apache.hadoop.hive.ql.lockmgr.DbTxnManager</value>
</property>
<property>
<name>hive.enforce.bucketing</name>
<value>true</value>
</property>
<property>
<name>hive.compactor.initiator.on</name>
<value>true</value>
</property>
<property>
<name>hive.compactor.worker.threads</name>
<value>2</value>
</property>
<property>
<name>hive.exec.dynamic.partition.mode</name>
<value>nonstrict</value>
</property>
<property>
<name>hive.compactor.initiator.on</name>
<value>true</value>
</property>
<!-- mysql 相关 -->
<property>
<name>javax.jdo.option.ConnectionURL</name>
<value>jdbc:mysql://localhost:3306/metastore?createDatabaseIfNotExist=true</value>
<description>Enter your JDBC connection string. </description>
</property>
<property>
<name>javax.jdo.option.ConnectionDriverName</name>
<value>com.mysql.jdbc.Driver</value>
</property>
<property>
<name>javax.jdo.option.ConnectionUserName</name>
<value>hive</value>
</property>
<property>
<name>javax.jdo.option.ConnectionPassword</name>
<value>hive</value>
</property>
<property>
<name>datanucleus.autoCreateSchema</name>
<value>false</value>
</property>
<property>
<name>datanucleus.fixedDatastore</name>
<value>true</value>
</property>
3. 创建具备ACID及Transactions的表
这里的表需要具备下面几个条件：
1. 必须以 ORC 格式存储
2. 必须分 bucket，且不能 sort
3. 必须显式声明transations
大致如下面模版所述：
create table T(...)
clustered by (a) into 2 buckets
stored as orc TBLPROPERTIES ('transactional'='true');知道了限定条件，我们来创建一个表做个测试：
create table goods_info (
shop_id int,
created_time string,
title string
)
clustered by (shop_id) into 3 buckets
stored as orc TBLPROPERTIES ('transactional'='true');之后insert两条数据
insert into table goods_info values (1, '2016-04-02 21:00:00', 'HIVE 7天入门'), (2, '2016-04-02 21:01:00', 'SICP');回车后，会运行一个MR去插入数据，之后就可以查询了
hive (default)> select * from goods_info;
OK
1 2016-04-02 21:00:00 HIVE 7)e�
2 2016-04-02 21:01:00 SICP
Time taken: 0.116 seconds, Fetched: 2 row(s)这里的中文貌似有乱码问题，这里先不做处理了。
在 hdfs 上生产的文件为：
210 /user/hive/warehouse/goods_info/delta_0000005_0000005/bucket_00000
810 /user/hive/warehouse/goods_info/delta_0000005_0000005/bucket_00001
781 /user/hive/warehouse/goods_info/delta_0000005_0000005/bucket_00002在创建表时，我遇到了下面的问题
FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. MetaException(message:javax.jdo.JDODataStoreException: An exception was thrown while adding/validating class(es) : Specified key was too long; max key length is 767 bytes
com.mysql.jdbc.exceptions.jdbc4.MySQLSyntaxErrorException: Specified key was too long; max key length is 767 bytes
at sun.reflect.GeneratedConstructorAccessor31.newInstance(Unknown Source)
at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
at java.lang.reflect.Constructor.newInstance(Constructor.java:422)
at com.mysql.jdbc.Util.handleNewInstance(Util.java:377)
at com.mysql.jdbc.Util.getInstance(Util.java:360)
at com.mysql.jdbc.SQLError.createSQLException(SQLError.java:978)
at com.mysql.jdbc.MysqlIO.checkErrorPacket(MysqlIO.java:3887)
at com.mysql.jdbc.MysqlIO.checkErrorPacket(MysqlIO.java:3823)
at com.mysql.jdbc.MysqlIO.sendCommand(MysqlIO.java:2435)
at com.mysql.jdbc.MysqlIO.sqlQueryDirect(MysqlIO.java:2582)
at com.mysql.jdbc.ConnectionImpl.execSQL(ConnectionImpl.java:2526)
at com.mysql.jdbc.ConnectionImpl.execSQL(ConnectionImpl.java:2484)
at com.mysql.jdbc.StatementImpl.execute(StatementImpl.java:848)
at com.mysql.jdbc.StatementImpl.execute(StatementImpl.java:742)
at com.jolbox.bonecp.StatementHandle.execute(StatementHandle.java:254)这时由于 mysql 自身对索引字段字符长度的限制，我数据库用的是utf8编码，767/3=255.6，这也就是说索引字段的varchar不能大于255，对于 mysql5.6以及以后的版本，可以通过在/etc/my.conf中添加如下配置解决
[mysqld]
innodb_file_format=BARRACUDA
innodb_large_prefix=on
innodb_file_per_table=true相关原理可见：
扫一扫
848
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
