结论:
a. Mysql 由于 next-lock 锁的存在,按更新语句的where条件的索引来锁定一个范围的记录。
如果索引不存在,那么会锁住整个表的记录。如果该条件是primary key,那么就是变为行锁。
b. next-lock是为了避免幻想读,所以在默认的 repeated read 的状况下会有这种情况。
c. 对于 read commited 的隔离级别,不会使用next-lock锁,而是使用行锁,这也是oracle的方式,
这种情况下,对并发的支持比 repeated read 要好些。
1. 设置会话级别的隔离级别为可重复读
set tx_isolation='repeatable-read';
2. 检查:会话级别已经修改,但是全局级别还是没有改的。
root@test 12:11:56>show variables like '%iso%';
+---------------+-----------------+
| Variable_name | Value |
+---------------+-----------------+
| tx_isolation | REPEATABLE-READ |
+---------------+-----------------+
1 row in set (0.00 sec)
root@test 12:12:06>show global variables like '%iso%';
+---------------+----------------+
| Variable_name | Value |
+---------------+----------------+
| tx_isolation | READ-COMMITTED |
+---------------+----------------+
3. 在该会话里开启一个事务,更新表的一行记录
root@test 12:19:12>begin;
Query OK, 0 rows affected (0.00 sec)
由于这个表里没有索引,所以由于next-lock的存在,所以即使只有更新一行,但是实际上已经对所有的记录上锁了。
root@test 12:19:20>update test set name='aa11' where id=1;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1 Changed: 1 Warnings: 0
4. 验证: 开启另一个会话,尝试更新另外一条数据,发现一直等锁,直到超时报错。
root@test 12:19:39>update test set name='bb11' where id=2;
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
insert 同样因为有锁,而发生等待,直到报错。
root@test 12:20:02>insert into test values (3,'cc');
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
5. 进行检查锁的信息:只有事务的信息,并没有锁的信息。即使在read-committed的隔离级别下也没有锁的信息。
select * from innodb_trx\G
*************************** 2. row ***************************
trx_id: 151933845
trx_state: RUNNING
trx_started: 2016-08-18 12:19:33
trx_requested_lock_id: NULL
trx_wait_started: NULL
trx_weight: 3
trx_mysql_thread_id: 31
trx_query: NULL
trx_operation_state: NULL
trx_tables_in_use: 0
trx_tables_locked: 0
trx_lock_structs: 2
trx_lock_memory_bytes: 360
trx_rows_locked: 3
trx_rows_modified: 1
trx_concurrency_tickets: 0
trx_isolation_level: REPEATABLE READ
trx_unique_checks: 1
trx_foreign_key_checks: 1
trx_last_foreign_key_error: NULL
trx_adaptive_hash_latched: 0
trx_adaptive_hash_timeout: 10000
trx_is_read_only: 0
trx_autocommit_non_locking: 0
2 rows in set (0.00 sec)
root@information_schema 12:31:26>select * from INNODB_LOCKS;
Empty set (0.00 sec)
状态信息Innodb_current_row_locks能显示出锁的信息。
root@test 01:43:13>show status like '%lock%';
+------------------------------------------+-------+
| Variable_name | Value |
+------------------------------------------+-------+
| Com_lock_tables | 0 |
| Com_lock_tables_for_backup | 0 |
| Com_lock_binlog_for_backup | 0 |
| Com_show_slave_status_nolock | 0 |
| Com_unlock_binlog | 0 |
| Com_unlock_tables | 0 |
| Handler_external_lock | 2 |
| Innodb_deadlocks | 0 |
| Innodb_row_lock_current_waits | 0 |
| Innodb_current_row_locks | 2 |
| Innodb_row_lock_time | 12002 |
| Innodb_row_lock_time_avg | 6001 |
| Innodb_row_lock_time_max | 6001 |
| Innodb_row_lock_waits | 2 |
| Innodb_s_lock_os_waits | 100 |
| Innodb_s_lock_spin_rounds | 3455 |
| Innodb_s_lock_spin_waits | 1550 |
| Innodb_x_lock_os_waits | 48 |
| Innodb_x_lock_spin_rounds | 5238 |
| Innodb_x_lock_spin_waits | 1539 |
| Key_blocks_not_flushed | 0 |
| Key_blocks_unused | 53578 |
| Key_blocks_used | 7 |
| Performance_schema_locker_lost | 0 |
| Performance_schema_rwlock_classes_lost | 0 |
| Performance_schema_rwlock_instances_lost | 0 |
| Qcache_free_blocks | 1 |
| Qcache_total_blocks | 1 |
| Table_locks_immediate | 126 |
| Table_locks_waited | 0 |
+------------------------------------------+-------+
30 rows in set (0.00 sec)
6. 进一步验证:INNODB_LOCKS 表的数据,只有等锁的语句正在执行的期间才会显示,Lock wait timeout 之前才会显示,这一点和Oracle的v$locked_object的视图信息是不一致的,
这个视图会显示当前哪些会话在哪些对象上上了锁。
root@information_schema 02:44:02>select * from INNODB_LOCKS \G
*************************** 1. row ***************************
lock_id: 151933860:250:3:1
lock_trx_id: 151933860
lock_mode: X
lock_type: RECORD
lock_table: `test`.`test`
lock_index: GEN_CLUST_INDEX
lock_space: 250
lock_page: 3
lock_rec: 1
lock_data: supremum pseudo-record
*************************** 2. row ***************************
lock_id: 151933845:250:3:1
lock_trx_id: 151933845
lock_mode: X
lock_type: RECORD
lock_table: `test`.`test`
lock_index: GEN_CLUST_INDEX
lock_space: 250
lock_page: 3
lock_rec: 1
lock_data: supremum pseudo-record
2 rows in set (0.00 sec)
--此时也能看出在等待那个锁,那个session
root@information_schema 04:41:40>select * from INNODB_LOCK_WAITS \G
*************************** 1. row ***************************
requesting_trx_id: 151933866 (请求锁的事务ID)e
requested_lock_id: 151933866:250:3:5
blocking_trx_id: 151933845 (阻塞锁的事务ID)
blocking_lock_id: 151933845:250:3:5
1 row in set (0.00 sec)
7. 综述如果有语句执行超时,要判断哪个线程,哪个语句锁住了数据,可以用下面的sql
select a.lock_table,a.lock_type,a.lock_space,c.trx_mysql_thread_id,
d.Command,d.Info, d.State,d.Time, d.Rows_sent, d.Rows_examined
from INNODB_LOCKS a, INNODB_LOCK_WAITS b, innodb_trx c, processlist d
where a.lock_trx_id=b.blocking_trx_id
and a.lock_trx_id=c.trx_id
and c.trx_mysql_thread_id=d.id
+---------------+-----------+------------+---------------------+---------+------+-------+-------+-----------+---------------+
| lock_table | lock_type | lock_space | trx_mysql_thread_id | Command | Info | State | Time | Rows_sent | Rows_examined |
+---------------+-----------+------------+---------------------+---------+------+-------+-------+-----------+---------------+
| `test`.`test` | RECORD | 250 | 31 | Sleep | NULL | | 10848 | 2 | 2 |
+---------------+-----------+------------+---------------------+---------+------+-------+-------+-----------+---------------+
1 row in set (0.00 sec)
至于这个线程执行的具体的sql语句,现在还没有发现在哪找得到,应该是slow sql 可以找到吧。
a. Mysql 由于 next-lock 锁的存在,按更新语句的where条件的索引来锁定一个范围的记录。
如果索引不存在,那么会锁住整个表的记录。如果该条件是primary key,那么就是变为行锁。
b. next-lock是为了避免幻想读,所以在默认的 repeated read 的状况下会有这种情况。
c. 对于 read commited 的隔离级别,不会使用next-lock锁,而是使用行锁,这也是oracle的方式,
这种情况下,对并发的支持比 repeated read 要好些。
1. 设置会话级别的隔离级别为可重复读
set tx_isolation='repeatable-read';
2. 检查:会话级别已经修改,但是全局级别还是没有改的。
root@test 12:11:56>show variables like '%iso%';
+---------------+-----------------+
| Variable_name | Value |
+---------------+-----------------+
| tx_isolation | REPEATABLE-READ |
+---------------+-----------------+
1 row in set (0.00 sec)
root@test 12:12:06>show global variables like '%iso%';
+---------------+----------------+
| Variable_name | Value |
+---------------+----------------+
| tx_isolation | READ-COMMITTED |
+---------------+----------------+
3. 在该会话里开启一个事务,更新表的一行记录
root@test 12:19:12>begin;
Query OK, 0 rows affected (0.00 sec)
由于这个表里没有索引,所以由于next-lock的存在,所以即使只有更新一行,但是实际上已经对所有的记录上锁了。
root@test 12:19:20>update test set name='aa11' where id=1;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1 Changed: 1 Warnings: 0
4. 验证: 开启另一个会话,尝试更新另外一条数据,发现一直等锁,直到超时报错。
root@test 12:19:39>update test set name='bb11' where id=2;
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
insert 同样因为有锁,而发生等待,直到报错。
root@test 12:20:02>insert into test values (3,'cc');
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
5. 进行检查锁的信息:只有事务的信息,并没有锁的信息。即使在read-committed的隔离级别下也没有锁的信息。
select * from innodb_trx\G
*************************** 2. row ***************************
trx_id: 151933845
trx_state: RUNNING
trx_started: 2016-08-18 12:19:33
trx_requested_lock_id: NULL
trx_wait_started: NULL
trx_weight: 3
trx_mysql_thread_id: 31
trx_query: NULL
trx_operation_state: NULL
trx_tables_in_use: 0
trx_tables_locked: 0
trx_lock_structs: 2
trx_lock_memory_bytes: 360
trx_rows_locked: 3
trx_rows_modified: 1
trx_concurrency_tickets: 0
trx_isolation_level: REPEATABLE READ
trx_unique_checks: 1
trx_foreign_key_checks: 1
trx_last_foreign_key_error: NULL
trx_adaptive_hash_latched: 0
trx_adaptive_hash_timeout: 10000
trx_is_read_only: 0
trx_autocommit_non_locking: 0
2 rows in set (0.00 sec)
root@information_schema 12:31:26>select * from INNODB_LOCKS;
Empty set (0.00 sec)
状态信息Innodb_current_row_locks能显示出锁的信息。
root@test 01:43:13>show status like '%lock%';
+------------------------------------------+-------+
| Variable_name | Value |
+------------------------------------------+-------+
| Com_lock_tables | 0 |
| Com_lock_tables_for_backup | 0 |
| Com_lock_binlog_for_backup | 0 |
| Com_show_slave_status_nolock | 0 |
| Com_unlock_binlog | 0 |
| Com_unlock_tables | 0 |
| Handler_external_lock | 2 |
| Innodb_deadlocks | 0 |
| Innodb_row_lock_current_waits | 0 |
| Innodb_current_row_locks | 2 |
| Innodb_row_lock_time | 12002 |
| Innodb_row_lock_time_avg | 6001 |
| Innodb_row_lock_time_max | 6001 |
| Innodb_row_lock_waits | 2 |
| Innodb_s_lock_os_waits | 100 |
| Innodb_s_lock_spin_rounds | 3455 |
| Innodb_s_lock_spin_waits | 1550 |
| Innodb_x_lock_os_waits | 48 |
| Innodb_x_lock_spin_rounds | 5238 |
| Innodb_x_lock_spin_waits | 1539 |
| Key_blocks_not_flushed | 0 |
| Key_blocks_unused | 53578 |
| Key_blocks_used | 7 |
| Performance_schema_locker_lost | 0 |
| Performance_schema_rwlock_classes_lost | 0 |
| Performance_schema_rwlock_instances_lost | 0 |
| Qcache_free_blocks | 1 |
| Qcache_total_blocks | 1 |
| Table_locks_immediate | 126 |
| Table_locks_waited | 0 |
+------------------------------------------+-------+
30 rows in set (0.00 sec)
6. 进一步验证:INNODB_LOCKS 表的数据,只有等锁的语句正在执行的期间才会显示,Lock wait timeout 之前才会显示,这一点和Oracle的v$locked_object的视图信息是不一致的,
这个视图会显示当前哪些会话在哪些对象上上了锁。
root@information_schema 02:44:02>select * from INNODB_LOCKS \G
*************************** 1. row ***************************
lock_id: 151933860:250:3:1
lock_trx_id: 151933860
lock_mode: X
lock_type: RECORD
lock_table: `test`.`test`
lock_index: GEN_CLUST_INDEX
lock_space: 250
lock_page: 3
lock_rec: 1
lock_data: supremum pseudo-record
*************************** 2. row ***************************
lock_id: 151933845:250:3:1
lock_trx_id: 151933845
lock_mode: X
lock_type: RECORD
lock_table: `test`.`test`
lock_index: GEN_CLUST_INDEX
lock_space: 250
lock_page: 3
lock_rec: 1
lock_data: supremum pseudo-record
2 rows in set (0.00 sec)
--此时也能看出在等待那个锁,那个session
root@information_schema 04:41:40>select * from INNODB_LOCK_WAITS \G
*************************** 1. row ***************************
requesting_trx_id: 151933866 (请求锁的事务ID)e
requested_lock_id: 151933866:250:3:5
blocking_trx_id: 151933845 (阻塞锁的事务ID)
blocking_lock_id: 151933845:250:3:5
1 row in set (0.00 sec)
7. 综述如果有语句执行超时,要判断哪个线程,哪个语句锁住了数据,可以用下面的sql
select a.lock_table,a.lock_type,a.lock_space,c.trx_mysql_thread_id,
d.Command,d.Info, d.State,d.Time, d.Rows_sent, d.Rows_examined
from INNODB_LOCKS a, INNODB_LOCK_WAITS b, innodb_trx c, processlist d
where a.lock_trx_id=b.blocking_trx_id
and a.lock_trx_id=c.trx_id
and c.trx_mysql_thread_id=d.id
+---------------+-----------+------------+---------------------+---------+------+-------+-------+-----------+---------------+
| lock_table | lock_type | lock_space | trx_mysql_thread_id | Command | Info | State | Time | Rows_sent | Rows_examined |
+---------------+-----------+------------+---------------------+---------+------+-------+-------+-----------+---------------+
| `test`.`test` | RECORD | 250 | 31 | Sleep | NULL | | 10848 | 2 | 2 |
+---------------+-----------+------------+---------------------+---------+------+-------+-------+-----------+---------------+
1 row in set (0.00 sec)
至于这个线程执行的具体的sql语句,现在还没有发现在哪找得到,应该是slow sql 可以找到吧。
来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/674865/viewspace-2123686/,如需转载,请注明出处,否则将追究法律责任。
转载于:http://blog.itpub.net/674865/viewspace-2123686/