关于事务超时 · k8s原理与实践

[TOC] # 关于事务超时 OceanBase 为了避免事务长时间不提交持有锁影响其他会话，设计了两个超时逻辑。一个是事务空闲超时，一个是事务未提交超时。分别由租户变量 ob\_trx\_idle\_timeout 和 ob\_trx\_timeout 控制，默认值分别是 120 秒和 100 秒。通常只会有一个超时机制被触发。 **说明** 建议变量 ob\_trx\_idle\_timeout 使用默认值。 ~~~ obclient> show variables where variable_name in ('ob_trx_idle_timeout','ob_trx_timeout'); +---------------------+-----------+ | VARIABLE_NAME | VALUE | +---------------------+-----------+ | ob_trx_idle_timeout | 120000000 | | ob_trx_timeout | 100000000 | +---------------------+-----------+ 2 rows in set (0.00 sec) ~~~ ## 关于事务空闲超时 OceanBase 的事务空闲时间超过一段时间还没有提交时，会自动断开连接并回滚事务，此时会话需要重新连接。会话事务空闲超时时间阈值由租户变量 ob\_trx\_idle\_timeout 控制，这个参数值建议使用默认值 120 秒，实际空闲会话断开的时间会是在 \[100s, 100s + ob\_trx\_idle\_timeout \] 之间。 #### 示例：事务空闲超时报错下面示例先设置事务空闲超时时间为 120 秒，事务未提交超时时间为 1000 秒。当事务空闲时间超过 120 秒后，连接会被自动断开，事务也自动被 ROLLBACK 了。 ~~~ obclient> DROP TABLE IF EXISTS t_insert; Query OK, 0 rows affected (0.01 sec) obclient> CREATE TABLE t_insert( id bigint NOT NULL PRIMARY KEY auto_increment , name varchar(10) NOT NULL , value bigint ,gmt_create timestamp NOT NULL DEFAULT current_timestamp ); Query OK, 0 rows affected (0.05 sec) obclient> INSERT INTO t_insert(name, value) values('CN',NULL),('UK',NULL),('US',NULL); Query OK, 3 rows affected (0.00 sec) Records: 3 Duplicates: 0 Warnings: 0 obclient> select now(), * from t_insert t; +---------------------+----+------+-------+---------------------+ | now() | id | name | value | gmt_create | +---------------------+----+------+-------+---------------------+ | 2020-04-03 16:54:51 | 1 | CN | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:54:51 | 2 | UK | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:54:51 | 3 | US | NULL | 2020-04-03 16:54:49 | +---------------------+----+------+-------+---------------------+ 3 rows in set (0.00 sec) obclient> set session ob_trx_timeout=1000000000; Query OK, 0 rows affected (0.00 sec) obclient> set session ob_trx_idle_timeout=120000000; Query OK, 0 rows affected (0.00 sec) obclient> begin; Query OK, 0 rows affected (0.00 sec) obclient> update t_insert set gmt_create=now() where id=3; Query OK, 1 row affected (0.00 sec) Rows matched: 1 Changed: 1 Warnings: 0 obclient> select now(), t.* from t_insert t; +---------------------+----+------+-------+---------------------+ | now() | id | name | value | gmt_create | +---------------------+----+------+-------+---------------------+ | 2020-04-03 16:55:30 | 1 | CN | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:55:30 | 2 | UK | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:55:30 | 3 | US | NULL | 2020-04-03 16:55:25 | +---------------------+----+------+-------+---------------------+ 3 rows in set (0.00 sec) <<等 100 秒不操作>> obclient> select now(), t.* from t_insert t; ERROR-02013: Lost connection to MySQL server during query obclient> select now(), * from t_insert t; ERROR-02006: MySQL server has gone away No connection. Trying to reconnect... Connection id: 53246 Current database: TPCC +---------------------+----+------+-------+---------------------+ | now() | id | name | value | gmt_create | +---------------------+----+------+-------+---------------------+ | 2020-04-03 16:57:41 | 1 | CN | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:57:41 | 2 | UK | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:57:41 | 3 | US | NULL | 2020-04-03 16:54:49 | +---------------------+----+------+-------+---------------------+ 3 rows in set (0.00 sec) ~~~ ## 关于事务未提交超时 OceanBase 的事务持续时间超过一段时间还没有提交，会报超时错误。此时会话需要明确发出 ROLLBACK 命令才可以继续在会话里执行 SQL。会话事务的未提交超时时间阈值是由租户变量 ob\_trx\_timeout 控制。 #### 示例：事务未提交超时报错下面示例先设置事务空闲超时时间为 120 秒，事务超时时间为 100 秒。当一个事务未提交时间持续到 100 秒时，事务内部状态就变为超时状态，同时锁会释放。此后会话需要显式发出 ROLLBACK 语句。 ~~~ obclient> set session ob_trx_timeout=100000000; Query OK, 0 rows affected (0.00 sec) obclient> set session ob_trx_idle_timeout=120000000; Query OK, 0 rows affected (0.00 sec) obclient> begin; Query OK, 0 rows affected (0.00 sec) obclient> update t_insert set gmt_create=sysdate() where id=3; Query OK, 1 row affected (0.00 sec) Rows matched: 1 Changed: 1 Warnings: 0 obclient> select now(), t.* from t_insert t ; +---------------------+----+------+-------+---------------------+ | now() | id | name | value | gmt_create | +---------------------+----+------+-------+---------------------+ | 2020-04-03 16:59:56 | 1 | CN | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:59:56 | 2 | UK | NULL | 2020-04-03 16:54:49 | | 2020-04-03 16:59:56 | 3 | US | NULL | 2020-04-03 16:59:51 | +---------------------+----+------+-------+---------------------+ 3 rows in set (0.00 sec) <<等 120 秒不操作>> obclient> select now(), t.* from t_insert t ; ERROR-00600: internal error code, arguments: -6210, Transaction is timeout obclient> commit; ERROR-00600: internal error code, arguments: -6210, Transaction is timeout obclient> rollback; Query OK, 0 rows affected (0.00 sec) obclient> select now(), t.* from t_insert t ; +---------------------+----+------+-------+---------------------+ | now() | id | name | value | gmt_create | +---------------------+----+------+-------+---------------------+ | 2020-04-03 17:04:13 | 1 | CN | NULL | 2020-04-03 16:54:49 | | 2020-04-03 17:04:13 | 2 | UK | NULL | 2020-04-03 16:54:49 | | 2020-04-03 17:04:13 | 3 | US | NULL | 2020-04-03 16:54:49 | +---------------------+----+------+-------+---------------------+ 3 rows in set (0.00 sec) ~~~ **说明** 建议不要将事务未提交超时参数设置小于 1 秒。