MySQL极慢的查询

Question

I am trying to learn about optimizations to MySQL, table engines and when to use them, etc.

I have a query that is running up against the time-out limit of 10 minutes and which needs to complete in seconds because its function is a user-generated report.

The Query:

SELECT em.employeeId, tsk.taskId 
  FROM employee em INNER JOIN 
                      task tsk 
                   ON tsk.employeeId = em.employeeId
WHERE em.employeeId <> 'Not Done' 
  AND tsk.employeeId (
                SELECT employeeId FROM task
                 WHERE templateId 
                    IN ( '5', '6', '7', '8' )
                   AND tsk.status = 'Done'
                )
AND tsk.employeeId IN
(
    SELECT employeeId FROM task
    WHERE templateId IN
                ( '55', '56', '57', '58' ) 
      AND status = 'Not Done'
)

Explain:

# id, select_type, table, type, possible_keys, key, key_len, ref, rows, Extra
1, PRIMARY, tsk, ALL, , , , , 61326, Using where
1, PRIMARY, em, eq_ref, PRIMARY, PRIMARY, 4, newhire.tsk.employeeId, 1, Using index
3, DEPENDENT SUBQUERY, task, ALL, , , , , 61326, Using where
2, DEPENDENT SUBQUERY, task, ALL, , , , , 61326, Using where

The DB server uses MyISAM as default, so most schemas including this one are MyISAM.

I also realize that the text searches ( status=Done or status LIKE 'Done' ) are adding a lot to the query.

EDIT1:

# Table, Create Table
employee, CREATE TABLE `employee` (
  `employeeId` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `lastName` varchar(255) NOT NULL,
  `firstName` varchar(255) NOT NULL,
  `applicantId` varchar(255) NOT NULL,
  `fEmployeeId` varchar(255) DEFAULT NULL,
  `opId` varchar(255) DEFAULT NULL,
  `rehire` tinyint(3) unsigned NOT NULL DEFAULT '0',
  `sDate` date DEFAULT NULL,
  `oDate` date DEFAULT NULL,
  `cDate` date DEFAULT NULL,
  `additionalDate` date DEFAULT NULL,
  `additionalType` varchar(255) DEFAULT NULL,
  `processingDate` date DEFAULT NULL,
  `created` datetime NOT NULL,
  `recruiterId` int(10) unsigned NOT NULL,
  `processorId` int(10) unsigned DEFAULT NULL,
  `position` tinyint(3) unsigned NOT NULL DEFAULT '1',
  `status` varchar(255) NOT NULL,
  `campus` varchar(255) DEFAULT NULL,
  `phone` varchar(255) DEFAULT NULL,
  `email` varchar(255) DEFAULT NULL,
  `requisition` varchar(255) DEFAULT NULL,
  `Position` varchar(255) DEFAULT NULL,
  `department` varchar(255) DEFAULT NULL,
  `jobClass` varchar(255) DEFAULT NULL,
  `hiringManager` varchar(255) DEFAULT NULL,
  `badge` varchar(255) DEFAULT NULL,
  `currentAddress` varchar(255) DEFAULT NULL,
  `holding` tinyint(3) unsigned DEFAULT '0',
  PRIMARY KEY (`employeeId`)
) ENGINE=MyISAM AUTO_INCREMENT=3959 DEFAULT CHARSET=latin1

EDIT 2:

# Table, Create Table
task, CREATE TABLE `task` (
  `taskId` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `templateId` int(10) unsigned NOT NULL,
  `employeeId` int(10) unsigned NOT NULL,
  `name` varchar(255) NOT NULL,
  `description` text,
  `naAvailable` tinyint(3) unsigned DEFAULT '0',
  `fileRequired` tinyint(3) unsigned DEFAULT '0',
  `fileHrCatalog` int(10) unsigned DEFAULT NULL,
  `quickFileName` varchar(255) DEFAULT NULL,
  `fileUploaded` tinyint(3) unsigned DEFAULT '0',
  `fileExt` varchar(255) DEFAULT NULL,
  `level` tinyint(3) unsigned NOT NULL,
  `status` varchar(255) NOT NULL,
  `due` date DEFAULT NULL,
  `daysDue` int(10) unsigned DEFAULT NULL,
  `routeIncentives` tinyint(3) unsigned DEFAULT '0',
  `requiresAudit` tinyint(3) unsigned DEFAULT '0',
  `auditStatus` varchar(255) DEFAULT NULL,
  `auditUser` int(10) unsigned DEFAULT NULL,
  `auditDate` datetime DEFAULT NULL,
  `stampOption` tinyint(3) unsigned DEFAULT '0',
  `done` tinyint(3) unsigned DEFAULT '0',
  `doneBy` int(10) unsigned DEFAULT NULL,
  `doneWhen` datetime DEFAULT NULL,
  `sortOrder` tinyint(3) unsigned NOT NULL DEFAULT '255',
  PRIMARY KEY (`taskId`),
  KEY `status` (`status`,`templateId`)
) ENGINE=MyISAM AUTO_INCREMENT=176802 DEFAULT CHARSET=latin1

Answer 1

I would write the query as below, but to help the optimization, have a covering indexes on your tables.

Employee table -- index on ( status, employeeID ) Task table -- index on ( employeeid, templateid, status )

By the first join, you are prequalifying to get the first task as a "Done" status.

The second join is looking for the OTHER task you are interested in that is NOT Done.

Doing subqueries (especially correlated sub queries) can be harder on performance. By doing a JOIN, it's either there or its not...

SELECT 
      em.employeeId, 
      tsk.taskId 
   FROM 
      employee em 
         INNER JOIN task tsk1
            ON em.employeeId = tsk1.employeeId
            AND tsk1.templateID in ( '5', '6', '7', '8' )
            AND tsk1.status = 'Done'

         INNER JOIN task tsk2
            ON em.employeeId = tsk2.employeeId
            AND tsk2.templateID in ( '55', '56', '57', '58' )
            AND tsk2.status = 'Not Done'
   WHERE 
     em.status <> 'Not Done' 

Answer 2

Your first change should be to create an index on task that covers both the status and templateId columns:

ALTER TABLE task ADD INDEX (status, templateId);

That'll prevent the full-table scans of 61326 rows each time that table is accessed in your query.

Also, it looks like you might have made a typo here:

SELECT employeeId FROM task
WHERE templateId 
IN ( '5', '6', '7', '8' )
AND tsk.status = 'Done'

That tsk.status should be just status like the 2nd subquery.