Question

I have a pretty big table with contains about 3 million records.
When running a very simple query, joining this table on a few others (all with indexes and/or primary keys), the query will take about 25 seconds to complete!
The value of “Handler_read_next” is about 7 million!

Number of requests to read the next row in key order, incremented if you are querying an index column with a range constraint or if you are doing an index scan.

This problem have only started since this table began to grow big.

Now if I do an “optimize tables” on this table, the query will run in about 0.02 seconds and “Handler_read_next” will have a value of about 1500.

How can the difference be so extreme, and do I really have to setup a scheduled query, optimizing this table once a week or so? Even so, I would like to know the meaning behind this and why mysql behaves like this. Sure, rows are deleted and updated pretty much in this table, but should it get so badly fragmented in only one week that the query goes from 0.02 sec to 25 sec?

Edit: After request, here comes the query in question:

SELECT *
FROM budget_expenses 
  JOIN budget_categories 
    ON  budget_categories.BudgetAreaId = budget_expenses.BudgetAreaId 
    AND budget_categories.BudgetCategoryId = budget_expenses.BudgetCategoryId
  LEFT JOIN budget_types 
    ON  budget_types.BudgetAreaId = budget_expenses.BudgetAreaId 
    AND budget_types.BudgetCategoryId = budget_expenses.BudgetCategoryId 
    AND budget_types.BudgetTypeId = budget_expenses.BudgetTypeId
WHERE budget_expenses.BudgetId = 1 
  AND budget_expenses.ExpenseDate >= '2012-11-25' 
  AND budget_expenses.ExpenseDate <= '2012-12-24' 
  AND budget_expenses.BudgetAreaId = 2 
ORDER BY budget_expenses.ExpenseDate DESC, 
         budget_expenses.ExpenseTime IS NULL ASC, 
         budget_expenses.ExpenseTime DESC

(BudgetAreaId, BudgetCategoryId) is the primary key in budget_categories and (BudgetAreaId, BudgetCategoryId, BudgetTypeId) is the primary key in budget_types. In budget_expenses these 3 keys are indexes and also ExpenseDate has an index. This query returns about 20 rows.

Show create table:

CREATE TABLE `budget_areas` (
  `BudgetAreaId` int(11) NOT NULL,
  `Name` varchar(255) DEFAULT NULL,
  PRIMARY KEY (`BudgetAreaId`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1

CREATE TABLE `budget_categories` (
  `BudgetAreaId` int(11) NOT NULL,
  `BudgetCategoryId` int(11) NOT NULL AUTO_INCREMENT,
  `Name` varchar(255) DEFAULT NULL,
  `SortOrder` int(11) DEFAULT NULL,
  PRIMARY KEY (`BudgetAreaId`,`BudgetCategoryId`),
  KEY `BudgetAreaId` (`BudgetAreaId`,`BudgetCategoryId`)
) ENGINE=MyISAM DEFAULT CHARSET=latin1


CREATE TABLE `budget_types` (
  `BudgetAreaId` int(11) NOT NULL,
  `BudgetCategoryId` int(11) NOT NULL,
  `BudgetTypeId` int(11) NOT NULL,
  `Name` varchar(255) DEFAULT NULL,
  `SortId` int(11) DEFAULT NULL,
  PRIMARY KEY (`BudgetAreaId`,`BudgetCategoryId`,`BudgetTypeId`),
  KEY `BudgetAreaId` (`BudgetAreaId`,`BudgetCategoryId`,`BudgetTypeId`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1

 CREATE TABLE `budget_expenses` (
  `ExpenseId` int(11) NOT NULL AUTO_INCREMENT,
  `BudgetId` int(11) NOT NULL,
  `TempId` int(11) DEFAULT NULL,
  `BudgetAreaId` int(11) DEFAULT NULL,
  `BudgetCategoryId` int(11) DEFAULT NULL,
  `BudgetTypeId` int(11) DEFAULT NULL,
  `Company` varchar(255) DEFAULT NULL,
  `ImportCompany` varchar(255) DEFAULT NULL,
  `Sum` double(50,2) DEFAULT NULL,
  `ExpenseDate` date DEFAULT NULL,
  `ExpenseTime` time DEFAULT NULL,
  `Inserted` datetime DEFAULT NULL,
  `Changed` datetime DEFAULT NULL,
  `InsertType` int(1) DEFAULT NULL,
  `AccountId` int(11) DEFAULT NULL,
  `BankCardId` int(11) DEFAULT NULL,
  PRIMARY KEY (`ExpenseId`),
  KEY `BudgetId` (`BudgetId`),
  KEY `AccountId` (`AccountId`),
  KEY `Company` (`Company`) USING BTREE,
  KEY `ExpenseDate` (`ExpenseDate`),
  KEY `BudgetAreaId` (`BudgetAreaId`),
  KEY `BudgetCategoryId` (`BudgetCategoryId`),
  KEY `BudgetTypeId` (`BudgetTypeId`),
  CONSTRAINT `budget_expenses_ibfk_1` FOREIGN KEY (`BudgetId`) REFERENCES `budgets`    (`BudgetId`)
) ENGINE=InnoDB AUTO_INCREMENT=3604462 DEFAULT CHARSET=latin1

After I copy pasted this I changed from MyIsam to Innodb on the budget_categories table.

Edit: The change from myisam to innodb didn’t make any difference. The query is now very slow, just 12 hours after i optimized the budget_expenses table!

Here is the explain for the query which now takes about 9 seconds:

http://jsfiddle.net/dmVPY/1/

Answer 1

Ahhh MyISAM….

Try changing the table type (aka ‘storage engine’) to InnoDB instead.

If you do this, make sure innodb_buffer_pool_size in your my.cnf is a sensible value – the default is too small.