Atomikos实现分布式事物管理

Posted 2020-08-14 一步一脚印

　　分布式事务管理，简单来说就是多个数据库之间的事务管理。之前也试过使用Spring动态切换数据库,通过继承AbstractRoutingDataSource重写determineCurrentLookupKey()方法，来决定使用那个数据库。在开启事务之前，通过改变lookupKey来达到切换数据源目的。但是这种方法多个数据库之前没办法做事务管理，或许对于主从（读写）数据库会好用一些，而对于需要一个操作中更新多个数据库的情况，使用Atomikos或许会更好一些。

　　本文采用spring4+hibernate4+Atomikos进行事务管理。

　　Atomikos要使用的库使用maven导入，spring和hibernate的依赖就不一一列出，需要额外添加的dependency如下：

　　　　<dependency>
            <groupId>com.atomikos</groupId>
            <artifactId>atomikos-util</artifactId>
            <version>3.9.3</version>
        </dependency>
        <dependency>
            <groupId>com.atomikos</groupId>
            <artifactId>transactions-jdbc</artifactId>
            <version>3.9.3</version>
        </dependency>
        <dependency>
            <groupId>com.atomikos</groupId>
            <artifactId>transactions-jta</artifactId>
            <version>3.9.3</version>
        </dependency>
        <dependency>
            <groupId>com.atomikos</groupId>
            <artifactId>transactions</artifactId>
            <version>3.9.3</version>
        </dependency>
        <dependency>
            <groupId>com.atomikos</groupId>
            <artifactId>transactions-api</artifactId>
            <version>3.9.3</version>
        </dependency>
        <dependency>
            <groupId>javax.transaction</groupId>
            <artifactId>jta</artifactId>
            <version>1.1</version>
        </dependency>

　　使用的数据库驱动依赖如下：

        <!-- 加入mysql驱动依赖包 -->
        <dependency>
            <groupId>mysql</groupId>
            <artifactId>mysql-connector-java</artifactId>
            <version>5.1.34</version>
        </dependency>
        <!-- sql server数据库驱动 -->
        <dependency>
            <groupId>net.sourceforge.jtds</groupId>
            <artifactId>jtds</artifactId>
            <version>1.3.1</version>
        </dependency>
        <!-- 加入druid数据源依赖包 -->
        <dependency>
            <groupId>com.alibaba</groupId>
            <artifactId>druid</artifactId>
            <version>1.0.26</version>
        </dependency>            

　　有点需要注意的是，druid的版本应该跟上mysql的版本，不然可能找不到MySQL的XADataSource。

　　配置数据源xml代码：

　　<bean id="mysql" class="com.alibaba.druid.pool.xa.DruidXADataSource" init-method="init" destroy-method="close">
        <property name="url" value="${jdbc_url}" />
        <property name="username" value="${jdbc_username}" />
        <property name="password" value="${jdbc_password}" />

        <!-- 初始化连接大小 -->
        <property name="initialSize" value="0" />
        <!-- 连接池最大使用连接数量 -->
        <property name="maxActive" value="20" />
        <!-- 连接池最小空闲 -->
        <property name="minIdle" value="0" />
        <!-- 获取连接最大等待时间 -->
        <property name="maxWait" value="60000" />
        <!-- <property name="poolPreparedStatements" value="true" />
        <property name="maxPoolPreparedStatementPerConnectionSize" value="33" /> -->
        <property name="validationQuery" value="SELECT 1" />
        <property name="testOnBorrow" value="false" />
        <property name="testOnReturn" value="false" />
        <property name="testWhileIdle" value="true" />
        <!-- 配置间隔多久才进行一次检测，检测需要关闭的空闲连接，单位是毫秒 -->
        <property name="timeBetweenEvictionRunsMillis" value="60000" />
        <!-- 配置一个连接在池中最小生存的时间，单位是毫秒 -->
        <property name="minEvictableIdleTimeMillis" value="25200000" />
        <!-- 打开removeAbandoned功能 -->
        <property name="removeAbandoned" value="true" />
        <!-- 1800秒，也就是30分钟 -->
        <property name="removeAbandonedTimeout" value="1800" />
        <!-- 关闭abanded连接时输出错误日志 -->
        <property name="logAbandoned" value="true" />

        <!-- 监控数据库 -->
        <!-- <property name="filters" value="mergeStat" /> -->
        <!-- <property name="filters" value="stat" /> -->
        <!--<property name="filters" value="config" />
          <property name="connectionProperties" value="config.decrypt=true" />-->
    </bean>

    <bean id="mysqlDataSource" class="com.atomikos.jdbc.AtomikosDataSourceBean"
          init-method="init" destroy-method="close">
        <!-- Set unique name for this DataSource -->
        <property name="uniqueResourceName">
            <value>mysql</value>
        </property>
        <!-- Set XADatasource class name-->
        <property name="xaDataSource" ref="mysql"/>
        <!-- set properties for datasource connection pool -->
        <property name="poolSize" value="3" />
        <!-- 管理 Connection 被占用的时间 -->
        <!-- 如果不设置这个值，Atomikos使用默认的300秒(即5分钟)，那么在处理大批量数据读取的时候，一旦超过5分钟，就会抛出类似 Resultset is close 的错误 -->
        <property name="reapTimeout"><value>20000</value></property>
    </bean>

　　配置SessionFactory的xml代码：

<bean id="noticesSessionFactory"
          class="org.springframework.orm.hibernate4.LocalSessionFactoryBean">
        <property name="dataSource" ref="mysqlDataSource" />
        <property name="jtaTransactionManager" ref="springTransactionManager"></property>
        <property name="packagesToScan">
            <list>
                <value>com.test.model</value>
            </list>
        </property>
        <property name="hibernateProperties">
            <props>
                <!-- 慎重(推荐禁止) 固定为update -->
                <prop key="hibernate.hbm2ddl.auto">${hibernate.hbm2ddl.auto}</prop>
                <prop key="hibernate.dialect">org.hibernate.dialect.MySQLDialect</prop>
                <prop key="hibernate.autoReconnect">true</prop>
                <prop key="hibernate.show_sql">${hibernate.show_sql}</prop>
                <prop key="hibernate.format_sql">${hibernate.format_sql}</prop>
                <prop key="hibernate.use_sql_comments">${hibernate.use_sql_comments}</prop>
                <prop key="hibernate.jdbc.batch_size">${hibernate.batch_size}</prop>
            </props>
        </property>
    </bean>

　　如果要使用getCurrentSession()的话，要注意加上：

<property name="jtaTransactionManager" ref="springTransactionManager"></property>

　　添加多个数据库的话，只需要把上面的xml再写一遍。

　　下面配置atomikos的事务管理器xml代码：

<!-- atomikos事务管理器 -->
    <bean id="atomikosTransactionManager" class="com.atomikos.icatch.jta.UserTransactionManager"
          init-method="init" destroy-method="close">
        <property name="forceShutdown">
            <value>true</value>
        </property>
    </bean>
    <bean id="atomikosUserTransaction" class="com.atomikos.icatch.jta.UserTransactionImp">
        <property name="transactionTimeout" value="300" />
    </bean>

　　spring的事务管理器xml代码：

<!-- spring 事务管理器 -->
    <bean id="springTransactionManager"
          class="org.springframework.transaction.jta.JtaTransactionManager">
        <property name="transactionManager" ref="atomikosTransactionManager"/>
        <property name="userTransaction" ref="atomikosUserTransaction" />
        <property name="allowCustomIsolationLevels" value="true"/>
    </bean>

    <!-- 注解方式配置事物-->
   <tx:annotation-driven transaction-manager="springTransactionManager" />

　　添加transactions.properties文件：

# SAMPLE PROPERTIES FILE FOR THE TRANSACTION SERVICE
# THIS FILE ILLUSTRATES THE DIFFERENT SETTINGS FOR THE TRANSACTION MANAGER
# UNCOMMENT THE ASSIGNMENTS TO OVERRIDE DEFAULT VALUES;

# Required: factory implementation class of the transaction core.
# NOTE: there is no default for this, so it MUST be specified! 
# 
com.atomikos.icatch.service=com.atomikos.icatch.standalone.UserTransactionServiceFactory

        
# Set base name of file where messages are output 
# (also known as the ‘console file‘).
#
# com.atomikos.icatch.console_file_name = tm.out

# Size limit (in bytes) for the console file;
# negative means unlimited.
#
# com.atomikos.icatch.console_file_limit=-1

# For size-limited console files, this option
# specifies a number of rotating files to 
# maintain.
#
# com.atomikos.icatch.console_file_count=1

# Set the number of log writes between checkpoints
#
# com.atomikos.icatch.checkpoint_interval=500

# Set output directory where console file and other files are to be put
# make sure this directory exists!
#
# com.atomikos.icatch.output_dir = ./

# Set directory of log files; make sure this directory exists!
#
# com.atomikos.icatch.log_base_dir = ./

# Set base name of log file
# this name will be  used as the first part of 
# the system-generated log file name
#
# com.atomikos.icatch.log_base_name = tmlog

# Set the max number of active local transactions 
# or -1 for unlimited.
#
# com.atomikos.icatch.max_actives = 50

# Set the default timeout (in milliseconds) for local transactions
#
# com.atomikos.icatch.default_jta_timeout = 10000

# Set the max timeout (in milliseconds) for local transactions
#
# com.atomikos.icatch.max_timeout = 300000

# The globally unique name of this transaction manager process
# override this value with a globally unique name
#
# com.atomikos.icatch.tm_unique_name = tm
    
# Do we want to use parallel subtransactions? JTA‘s default
# is NO for J2EE compatibility
#
#com.atomikos.icatch.serial_jta_transactions=false
                    
# If you want to do explicit resource registration then
# you need to set this value to false.
#
# com.atomikos.icatch.automatic_resource_registration=true  
    
# Set this to WARN, INFO or DEBUG to control the granularity
# of output to the console file.
#
# com.atomikos.icatch.console_log_level=WARN
    
# Do you want transaction logging to be enabled or not?
# If set to false, then no logging overhead will be done
# at the risk of losing data after restart or crash.
#
# com.atomikos.icatch.enable_logging=true

# Should two-phase commit be done in (multi-)threaded mode or not?
# Set this to false if you want commits to be ordered according
# to the order in which resources are added to the transaction.
#
# NOTE: threads are reused on JDK 1.5 or higher. 
# For JDK 1.4, thread reuse is enabled as soon as the 
# concurrent backport is in the classpath - see 
# http://mirrors.ibiblio.org/pub/mirrors/maven2/backport-util-concurrent/backport-util-concurrent/
#
# com.atomikos.icatch.threaded_2pc=false

# Should shutdown of the VM trigger shutdown of the transaction core too?
#
# com.atomikos.icatch.force_shutdown_on_vm_exit=false

　　好了，Atomikos算是配置完成，我们可以写代码测试一下。测试的代码这里就不写出了。

　　性能优化：尽管这个软件有着很大的优势，但是想要更好的发挥其作用，可以按以下的方法优化：

　　l 更高的内存，意味着更高的吞吐量（每秒的事务数目）

　　l 使连接池尽可能的大

　　l 一旦你不需要的连接请马上关闭它们。不要把你的应用程序放在缓存里，让内部连接池为你做这些，这将促使更高效的连接使用

　　l 不要让活动的事务闲置：终止所有情况下的事务，尤其是在异常报错情况下的事务。这将减少数据库的锁定时间，并且最大效率的处理启用的使用。

　　这么长的配置文件要小心点写，我的粗心结果就是datasource乱套了，导致花了不少时间排错。当然还有各种乱七八糟的糟心问题，这些问题会在后面的文章中列出来，大家可以注意避免。最后用上Atomikos的分布式事务，没有了各种的脏数据，舒心了不少。