innodb 锁分裂继承与迁移
innodb行锁简介
- 行锁类型
LOCK_S:共享锁
LOCK_X: 排他锁
- GAP类型
LOCK_GAP:只锁间隙
LOCK_REC_NO_GAP:只锁记录
LOCK_ORDINARY: 锁记录和记录之前的间隙
LOCK_INSERT_INTENTION: 插入意向锁,用于insert时检查锁冲突
每个行锁由锁类型和GAP类型组成
例如:
LOCK_X|LOCK_ORDINARY 表示对记录和记录之前的间隙加排他锁
LOCK_S|LOCK_GAP 表示只对记录前的间隙加共享锁
锁的兼容性:
值得注意的是,持有GAP的锁(LOCK_GAP和LOCK_ORDINARY)与其他非LOCK_INSERT_INTENTION的锁都是兼容的,也就是说,GAP锁就是为了防止插入的。
详细可以参考之前的月报
innodb 锁分裂、继承与迁移
这里的锁分裂和合并,只是针对innodb行锁而言的,而且一般只作用于GAP类型的锁。
-
锁分裂
插入的记录的间隙存在GAP锁,此时此GAP需分裂为两个GAP
lock_rec_inherit_to_gap_if_gap_lock:
for (lock = lock_rec_get_first(block, heap_no);
lock != NULL;
lock = lock_rec_get_next(heap_no, lock)) {
if (!lock_rec_get_insert_intention(lock)
&& (heap_no == PAGE_HEAP_NO_SUPREMUM
|| !lock_rec_get_rec_not_gap(lock))) {
lock_rec_add_to_queue(
LOCK_REC | LOCK_GAP | lock_get_mode(lock),
block, heir_heap_no, lock->index,
lock->trx, FALSE);
}
}
-
锁继承
删除的记录前存在GAP锁,此GAP锁会继承到要删除记录的下一条记录上
lock_rec_inherit_to_gap:
for (lock = lock_rec_get_first(block, heap_no);
lock != NULL;
lock = lock_rec_get_next(heap_no, lock)) {
if (!lock_rec_get_insert_intention(lock)
&& !((srv_locks_unsafe_for_binlog
|| lock->trx->isolation_level
<= TRX_ISO_READ_COMMITTED)
&& lock_get_mode(lock) ==
(lock->trx->duplicates ? LOCK_S : LOCK_X))) {
lock_rec_add_to_queue(
LOCK_REC | LOCK_GAP | lock_get_mode(lock),
heir_block, heir_heap_no, lock->index,
lock->trx, FALSE);
}
}
-
锁迁移
B数结构变化,锁信息也会随之迁移. 锁迁移过程中也涉及锁继承。
锁分裂示例
- 锁分裂例子
set global tx_isolation='repeatable-read';
create table t1(c1 int primary key, c2 int unique) engine=innodb;
insert into t1 values(1,1);
begin;
# supremum 记录上加 LOCK_X|LOCK_GAP 锁住(1~)
select * from t1 where c2=2 for update;
# 发现插入(3,3)的间隙存在GAP锁,因此给(3,3)加LOCK_X|LOCK_GAP锁。这样依然锁住了(1~)
insert into t1 values(3,3);
这里如果插入(3,3)没有给(3,3)加LOCK_X|LOCK_GAP,那么其他连接插入(2,2)就可以成功
锁继承示例
- 隔离级别repeatable-read
===== RR =====
set global tx_isolation='repeatable-read';
create table t1(c1 int primary key, c2 int unique) engine=innodb;
insert into t1 values(1,1),(2,2);
#会话信息
session 1: | session 2:
begin; |
#(1,1) 加LOCK_X|LOCK_REC_NOT_GAP |
delete from t1 where c1=1; |
|
| begin;
| # (1,1)加LOCK_X|LOCK_ORDINARY 等待
| select * from t1 where c1 <= 1 for update;
commit; |
| #(1,1)被删除,purge清理delete mark时,(1,1)上的锁继承到(2,2)上,锁为LOCK_X|LOCK_GAP
| #同时(1,1)上的锁都释放,session 2等待成功
验证:session 1执行insert into t1 values(1,1)发生了锁等待,说明(2,2)上有gap锁
mysql> select * from information_schema.innodb_locks;
+------------------------+-------------+-----------+-----------+-----------------+------------+------------+-----------+----------+-----------+
| lock_id | lock_trx_id | lock_mode | lock_type | lock_table | lock_index | lock_space | lock_page | lock_rec | lock_data |
+------------------------+-------------+-----------+-----------+-----------------+------------+------------+-----------+----------+-----------+
| 16582717714:888654:4:3 | 16582717714 | X,GAP | RECORD | `cleaneye`.`t1` | c2 | 888654 | 4 | 3 | 2 |
| 16582692183:888654:4:3 | 16582692183 | X,GAP | RECORD | `cleaneye`.`t1` | c2 | 888654 | 4 | 3 | 2 |
+------------------------+-------------+-----------+-----------+-----------------+------------+------------+-----------+----------+-----------+
2 rows in set (0.01 sec)
其中session 2 在(2,2) 加了LOCK_X|LOCK_GAP
session 1 在(2,2) 加了LOCK_X|LOCK_GAP|LOCK_INSERT_INTENTION. LOCK_INSERT_INTENTION与LOCK_GAP冲突发生等待
- 隔离级别read-committed
===== RC =====
set global tx_isolation='read-committed';
drop table t1;
create table t1(c1 int primary key) engine=innodb;
insert into t1 values(1),(2);
#会话信息
session 1 | session 2
begin; |
#(1) 加LOCK_X|LOCK_REC_NOT_GAP |
delete from t1 where c1=1; |
|
| begin;
| #(1)加LOCK_S|LOCK_REC_NOT_GAP 等待
| select *from t1 where c1 <=1 lock in share mode;
|
COMMIT: |
| #(1)被删除,purge清理delete mark时,(1)上的锁继承到(2)上,锁为LOCK_S|LOCK_GAP
| # 同时(1)上的锁都释放,session 2等待成功
|
验证
session 1执行insert into t1 values(1)发生了锁等待,说明(2)上有gap锁
mysql> select * from information_schema.innodb_locks;
+------------------------+-----------------+-----------+-----------+-------------+------------+------------+-----------+----------+-----------+
| lock_id | lock_trx_id | lock_mode | lock_type | lock_table | lock_index | lock_space | lock_page | lock_rec | lock_data |
+------------------------+-----------------+-----------+-----------+-------------+------------+------------+-----------+----------+-----------+
| 1705:32:3:3 | 1705 | X,GAP | RECORD | `test`.`t1` | PRIMARY | 32 | 3 | 3 | 2 |
| 421590768578232:32:3:3 | 421590768578232 | S,GAP | RECORD | `test`.`t1` | PRIMARY | 32 | 3 | 3 | 2 |
+------------------------+-----------------+-----------+-----------+-------------+------------+------------+-----------+----------+-----------+
X.GAP insert 加锁LOCK_X | LOCK_GAP | LOCK_INSERT_INTENTION
S.GAP 加锁LOCK_S|LOCK_GAP,记录(2)从删除的记录(1)继承过来的GAP锁
而实际在读提交隔离级别上,insert into t1 values(1)应该可以插入成功,不需要等待的,这个锁是否继承值得商榷。
来看一个插入成功的例子
===== RC =====
set global tx_isolation='read-committed';
drop table t1;
create table t1(c1 int primary key) engine=innodb;
insert into t1 values(1),(2);
# 会话信息
session 1 | session 2
|
| begin;
| #(1)加LOCK_S|LOCK_REC_NOT_GAP
| # 查询结果为(1,2)
| select *from t1 where c1 <=1 lock in share mode;
|
begin; |
# 检查(1)上的锁与LOCK_X|LOCK_GAP|LOCK_INSERT_INTENTION |
# 不冲突,插入成功 |
insert into t1 values(0); |
| #再次查询结果为(0,1,2)
commit; | select *from t1 where c1 <=3 lock in share mode;
- 隔离级别serializable
===== SERIALIZABLE =====
set global tx_isolation='SERIALIZABLE';
drop table t1;
create table t1(c1 int primary key) engine=innodb;
insert into t1 values(1),(2);
# 会话信息
session 1: | session 2:
begin; |
#(1) 加LOCK_X|LOCK_REC_NOT_GAP |
delete from t1 where c1=1; |
|
| begin;
| #(1)上加LOCK_S|LOCK_ORDINARY 等待
| select *from t1 where c1 <=1 ;
|
commit; |
|
| #(1)被删除,purge清理delete mark时,(1)上的锁继承到(2)上,锁为LOCK_S|LOCK_GAP
| # 同时(1)上的锁都释放,session 2等待成功
|
验证方法同read-committed。
B树结构变化与锁迁移
B树节点发生分裂,合并,删除都会引发锁的变化。锁迁移的原则是,B数结构变化前后,锁住的范围保证不变。
我们通过例子来说明
-
节点分裂
假设原节点A(infimum,1,3,supremum) 向右分裂为B(infimum,1,supremum), C(infimum,3,supremum)两个节点
infimum为节点中虚拟的最小记录,supremum为节点中虚拟的最大记录
假设原节点A上锁为3上LOCK_S|LOCK_ORIDNARY,supremum为LOCK_S|LOCK_GAP,实际锁住了(1~)
锁迁移过程大致为:1)将3上的gap锁迁移到C节点3上
2)将A上supremum迁移继承到C的supremum上
3)将C上最小记录3的锁迁移继承到B的supremum上
迁移完成后锁的情况如下(lock_update_split_right)
B节点:suprmum LOCK_S|LOCK_GAP
C节点:3 LOCK_S|LOCK_ORINARY, suprmum LOCK_S|GAP迁移后仍然锁住了范围(1~)
节点向左分裂情形类似
-
节点合并
以上述节点分裂的逆操作来讲述合并过程
B(infimum,1,supremum), C(infimum,3,supremum)两个节点,向左合并为A节点(infimum,1,3,supremum)
其中B,C节点锁情况如下
B节点:suprmum LOCK_S|LOCK_GAP
C节点:3 LOCK_S|LOCK_ORINARY, suprmum LOCK_S|GAP迁移流程如下(lock_update_merge_left):
1)将C节点锁记录3迁移到B节点
2)将B节点supremum迁移继承到A的supremum上
迁移后仍然锁住了范围(1~)
节点向右合并情形类似
-
节点删除
如果删除节点存在左节点,则将删除节点符合条件的锁,迁移继承到左节点supremum上
否则将删除节点符合条件的锁,迁移继承到右节点最小用户记录上
参考lock_update_discard
锁继承相关的BUG
bug#73170 二级唯一索引失效。这个bug触发条件是删除的记录没有被purge, 锁还没有被继承的。如果锁继承了就不会出现问题。
bug#76927 同样是二级唯一索引失效。这个bug是锁继承机制出了问题。
以上两个bug详情参考这里