Is there any way to reset mysql IDLE/Sleep connections? - mysql

Question

I have a problem, my app have to open a lot of connections with MySQL and write data into it, but for after some time I'm getting more and more connections that are in idle/sleep state, and my app's speed reduces until it completely stops. I blame 'wait_timeout' variable to be too high,
mysql> show variables like '%timeout%';
+----------------------------+----------+
| Variable_name | Value |
+----------------------------+----------+
| connect_timeout | 10 |
| delayed_insert_timeout | 300 |
| innodb_lock_wait_timeout | 50 |
| innodb_rollback_on_timeout | OFF |
| interactive_timeout | 28800 |
| lock_wait_timeout | 31536000 |
| net_read_timeout | 30 |
| net_write_timeout | 60 |
| slave_net_timeout | 3600 |
| wait_timeout | 28800 |
+----------------------------+----------+
10 rows in set (0.05 sec)
and
mysql> SHOW FULL PROCESSLIST;
+------+----------+----------------------------------------------------+-------------+---------+------+-------+-----------------------+
| Id | User | Host | db | Command | Time | State | Info |
+------+----------+----------------------------------------------------+-------------+---------+------+-------+-----------------------+
| 5425 | root | ip-10-xxxxxxxx.ec2.internal:60544 | my_db | Sleep | 2344 | | NULL |
| 5426 | root | ip-10-xxxxxxxx.ec2.internal:60561 | my_db | Sleep | 2136 | | NULL |
And there are a lot of connections like this.
So my primary question is how to reset connection? I know I can just kill PID but I would really like to reset it.
And do you think wait_timeout would solve an issue. By the way everything is on the Amazon RDS.

Answer 1

I would suggest that wait_timeout is probably the wrong solution, because, as #datasage implied, your application may not handle it gracefully. The solution is for your application to properly close the connections when it is finished with them.
KILL thread_id is the correct way to reset it from the server side, but that's just a patch for the actual solution, assuming this is actually what's slowing down your application.
I commonly have hundreds of sleeping threads on my MySQL servers, because the application holds them in a pool for reuse. Sleeping threads are sleeping. They aren't slowing anything down unless you have so many of them that you're hitting the max_connections limit on RDS.

Answer 2

You could reduce wait_timeout but you will need to be confident that your application can handle a connection being closed by the server instead of the client.