Mysql – Connecting to thesql with SUPER user when reached Too many connections

See here: http://bugs.mysql.com/bug.php?id=44030

Philip S

I had the same issue. I tried upgrading from 5.1.33 to 5.1.36, but mysqld would still crash whenever accessing the database.

Some details, most of it probably irrelevant, but maybe something will pop out: This is a brand new laptop, and I just started on a new project. MD Turion X2 Dual-Core Mobile RM-74 w/4GB RAM DB size on disk is 1.67GB 1GB of RAM is being used as a ramdisk Fresh install of 5.1.33 on MS Vista 64 using the WAMP project installer. A lot of ALTERs, as this is a dev environment. Adding columns, indexes, etc. Renaming columns as well. The bug cropped up between coding sessions rather than during a coding session. All db schema was created using MySQL Administrator, and all data manipulation done using php and mysqli. Everything wrt MySQL is brand new. I'm sorry, I'm unable to provide data files.

Anyway, as the upgrade didn't work to allow me to access the data, and a lot of work occurred since my last backup, I wanted to restore/retrieve as much of the DB schema as possible, I wasn't concerned with data. I actually did retrieve most of the data as well. While I'm sure the process can be optimized, what worked for me is: C:\wamp\bin\mysql\mysql5.1.36\bin>mysqlshow.exe -u root then for each table C:\wamp\bin\mysql\mysql5.1.36\bin>mysqlshow.exe -u root until I got the error mysqlshow.exe: Cannot list columns in db: , table: : Lost connection to MySQL server during query then I restarted mysqld, started the MySQL console, and \u drop table for the table which caused the crash. After that, everything was back to normal other than the loss of the one small table which I should be able to recreate easily.

There are several solutions.

To get god insert performance with minimal impact on your code, have a look at mysql clusters. These go well beyond replication and transparently implement sharding. I believe (but would need to dig about to verify) that a mysql cluster can act as a master in master/slave replication. So, eg. you might have a 4 node cluster handling writes replicating to a dozen or so slaves.

Note that you can implement master/master replication - you can effectively have any number of nodes arranged in a ring - which will also give you the benefit of insert performance - but with such a large number of nodes, there is increased risk of delays in propagation of updates.

If you've got a complex schema, then you might get big benefits by using the federated storage engine to split the data, although mysql does not always optimize queries as much as it could in this scenario.

You should definitely be looking at mysqlproxy or some other abstraction layer if you're going down most of these routes.