为什么此SQL查询会杀死我的服务器？

Question

When I run this query on SQL server from a Varnish cache server.

select count(distinct email) from aiki_users, ocal_files where userid = 
       (select upload_user from ocal_files where upload_user = userid limit 1)

With aiki_users table having > 80000 rows

and the ocal_files table having > 30000 rows.

It kills the server. It is called from a php script via ajax. I have to close the page that calls it.

I don't know what it returns, but I know that the thing I want to do should be created with group by and having, but I don't know how.

The aiki_users table looks like this:

CREATE TABLE IF NOT EXISTS `aiki_users` (
  `userid` int(9) unsigned NOT NULL auto_increment,
  `username` varchar(100) NOT NULL default '',
  `full_name` varchar(255) NOT NULL,
  `country` varchar(255) NOT NULL,
  `sex` varchar(25) NOT NULL,
  `job` varchar(255) NOT NULL,
  `password` varchar(100) NOT NULL default '',
  `oauth_passcode` varchar(255) NOT NULL,
  `oauth_token` varchar(255) NOT NULL,
  `oauth_token_secret` varchar(255) NOT NULL,
  `usergroup` int(10) NOT NULL default '0',
  `email` varchar(100) NOT NULL default '',
  `avatar` varchar(255) NOT NULL,
  `homepage` varchar(100) NOT NULL default '',
  `first_ip` varchar(40) NOT NULL default '0',
  `first_login` datetime NOT NULL,
  `last_login` datetime NOT NULL,
  `last_ip` varchar(40) NOT NULL,
  `user_permissions` text NOT NULL,
  `maillist` int(1) NOT NULL,
  `logins_number` int(11) NOT NULL,
  `randkey` varchar(255) NOT NULL,
  `is_active` int(5) NOT NULL,
  PRIMARY KEY  (`userid`),
  KEY `username` (`username`)
) ENGINE=MyISAM  DEFAULT CHARSET=utf8 AUTO_INCREMENT=3 ;

I don't have correct access to the server, I can't see the defnition of ocal_files but:

 ocal_files.aiki_users = aiki_users.userid

Can anybody tell me how bad the query is and what it returns?

Answer 1

您可以从主查询中的aiki_user和ocal_files中进行选择，而无需指定join子句，嵌套查询中的where不会停止全表连接...

Answer 2

I'd recommend running EXPLAIN PLAN on both queries and seeing what the database tells you.

My guess is that you have to do a TABLE SCAN, so the performance is terrible.