时间:2021-09-02 10:08:45 | 栏目:Mysql | 点击:次
序列化的实现
InnoDB对于序列化的实现方式,是通过两种方式实现的。
第一种,当SELECT语句在一个显式的事务块内,如执行表11-9中的编号为1的情况,将施加LOCK_S锁,根据表11-6(记录锁事务锁相容表)可知,LOCK_S锁排斥写锁,所以序列化隔离级别下只允许并发地读取操作,并发写被禁止,因此实现了可序列化。
相应代码如下:
ha_innobase::external_lock(...) {... if (lock_type != F_UNLCK) { /* MySQL is setting a new table lock */ ... if (trx->isolation_level == TRX_ISO_SERIALIZABLE //序列化隔离级别 && m_prebuilt->select_lock_type == LOCK_NONE && thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { //且在一个显式事务块内部 /* To get serializable execution, we let InnoDB conceptually add 'LOCK IN SHARE MODE' to all SELECTs which otherwise would have been consistent reads. An exception is consistent reads in the AUTOCOMMIT=1 mode: we know that they are read-only transactions, and they can be serialized also if performed as consistent reads. */ m_prebuilt->select_lock_type = LOCK_S; //加读锁,即 'LOCK IN SHARE MODE' m_prebuilt->stored_select_lock_type = LOCK_S; } //否则,不加锁(这一点也很重要) ... } else { TrxInInnoDB::end_stmt(trx); DEBUG_SYNC_C("ha_innobase_end_statement"); } ...}
第二种,当SELECT语句不在一个显式的事务块内,则通过获取最新快照(在事务开始的时候,),然后读取数据。此时,因基于快照的一致性读不需要加锁,所以其加锁情况对应到了表11-9中的编号2对应的情况。
表11-9 序列化隔离级别加锁情况
说明:
S0:SELECT * FROM bluesea WHERE c1=2;
//使用主键索引做WHERE条件
另外,对于FLUSH...WITH READ LOCK语句,序列化隔离级别下也需要加读锁LOCK_S
代码如下:
ha_innobase::store_lock( ... /* Check for FLUSH TABLES ... WITH READ LOCK */ if (trx->isolation_level == TRX_ISO_SERIALIZABLE) { m_prebuilt->select_lock_type = LOCK_S; m_prebuilt->stored_select_lock_type = LOCK_S; } else { m_prebuilt->select_lock_type = LOCK_NONE; m_prebuilt->stored_select_lock_type = LOCK_NONE; } ... }
与序列化相关的,还有innobase_query_caching_of_table_permitted()
函数,序列化隔离级别不允许缓冲查询。
总结