Tddl的分布式Sequence

今天药忘吃喽~ 2022-10-01 15:53 142阅读 0赞

[2019独角兽企业重金招聘Python工程师标准>>> ][2019_Python_] ![hot3.png][]

## 概述 ##

*  在GroupSequence调用nextValue()时，**判定当前GroupSequenceDao是否存在缓存值currentRange**；  
    有则直接从currentRange通过 AtomicLong并发控制取nextValue.  
    没有： 以ReentrantLock方式通过GroupSequenceDao的nextRange取指定sequence-name的数据库表配置;  再generateNewValue： newValue =  oldValue + outStep ;  通过oldValue做乐观锁将新值newValue 更新进库表（update sequence set value = ?, gmt\_modified = ? where name = ? and value = ?）；
 *  **GroupSequenceDao的innerStep 与outStep 很重要！**  
    outStep ： 控制更新数据库sequnece的Value增长步长。 在GroupSequenceDao.init时**：**   outStep =innerStep \* dscount;  
    innerStep ：控制缓存的最小值与最大值的范围大小。 currentRange 的范围是\[newValue +1, newValue + innerStep \]

## 使用方式 ##

代码：

@Bean(name = "errorCenterIdSequence", initMethod = "init")
        public GroupSequence errorCenterIdSequence(SequenceDao sequenceDao) {
            GroupSequence sequence = new GroupSequence();
            sequence.setSequenceDao(sequenceDao);
            sequence.setName("seq_decision_exec_error_center_id");
            groupSequenceMap.putIfAbsent("seq_decision_exec_error_center_id", sequence);
            return sequence;
        }
    
        @Bean(initMethod = "init")
        public GroupSequenceDao idSequenceDao() {
            GroupSequenceDao idSequenceDao = new GroupSequenceDao();
            idSequenceDao.setAppName(appname);
            idSequenceDao.setDbGroupKeys(Lists.newArrayList(seqGroup));
            /*
             * 默认是sequence name value gmt_modified 
             * idSequenceDao.setDscount(2);
             * idSequenceDao.setInnerStep(1000); 
             * idSequenceDao.setRetryTimes(2);
             * idSequenceDao.setTableName("sequence");
             * idSequenceDao.setNameColumnName("name");
             * idSequenceDao.setValueColumnName("value");
             * idSequenceDao.setGmtModifiedColumnName("gmt_modified");
             */
            idSequenceDao.setAdjust(true);
            return idSequenceDao;
        }

*   GroupSequenceDao.setAdjust(true);  设置的这个"adjust" 会在GroupSequence的初始化时通过GroupSequenceDao的adjust方法校验sequence的name配置在数据库表中，并判定是否要自动插入表。
 *  GroupSequenceDao 约定了sequence表名和表字段； 且在init()时， dscount = max(dscount, 分库数)； 步长outStep  = innerStep \* dscount 。  
    ![85aebf94a1d0dd97d4bf5e6cbd51319ea4a.jpg][]

数据库表：

CREATE TABLE `sequence` (
      `id` bigint(20) unsigned NOT NULL AUTO_INCREMENT,
      `gmt_created` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
      `gmt_modified` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '修改时间',
      `name` varchar(64) DEFAULT NULL,
      `value` bigint(20) unsigned DEFAULT NULL,
      PRIMARY KEY (`id`),
      UNIQUE KEY `uk_sequence_name` (`name`)
    ) ENGINE=InnoDB AUTO_INCREMENT=100016 DEFAULT CHARSET=utf8 COMMENT='sequence生成器';

![03f2d1952a4b48dfff017f8fd3d9641c62b.jpg][]

转载于:https://my.oschina.net/u/3434392/blog/3021263

[2019_Python_]: https://my.oschina.net/u/2663968/blog/3061697
[hot3.png]: /images/20220113/b6dc666652ef4033944ae05d8e24f971.png
[85aebf94a1d0dd97d4bf5e6cbd51319ea4a.jpg]: /images/20220113/732dec9e912b40e6bd6482150b54b420.png
[03f2d1952a4b48dfff017f8fd3d9641c62b.jpg]: /images/20220113/1077cf33881944778f57c7a67594a294.png