具有限制的MySQL查询执行计划根据

Question

我有一个限制的查询，根据close所在的值提供截然不同的执行时间。我正在使用MySQL 5.1，在我最大的表格中查看大约100,000,000条记录＆＃34;警报＆＃34;。

当我看到这个执行计划时，我很高兴，我的查询不到一秒钟：

mysql> EXPLAIN SELECT alert.id, category.severity, category.classification, category.completion, alert.description,
->        alert.detectTime, analyzer.name, analyzer.manufacturer, analyzer.model, analyzer.version,
->        analyzer.class, analyzer.osType, analyzer.osVersion, analyzerNode.hostName,
->        analyzerNode.address, sourceNode.hostName, sourceNode.address, targetNode.hostName,
->        targetNode.address, sourceUser.userName, sourceUser.uid, targetUser.userName, targetUser.uid,
->        alert.filePath, alert.sourceProcessPid, alert.sourceProcessPath, alert.targetProcessPid,
->        alert.targetProcessPath, alert.acknowledgeUserName, alert.acknowledgeTime, reference.url, reference.meaning
-> FROM alerts AS alert
->      LEFT JOIN categories AS category ON category.id = alert.categoryId
->      LEFT JOIN analyzers AS analyzer ON analyzer.id = alert.analyzerId
->      LEFT JOIN nodes AS analyzerNode ON analyzerNode.id = alert.analyzerNodeId
->      LEFT JOIN nodes AS sourceNode ON sourceNode.id = alert.sourceNodeId
->      LEFT JOIN nodes AS targetNode ON targetNode.id = alert.targetNodeId
->      LEFT JOIN users AS sourceUser ON sourceUser.id = alert.sourceUserId
->      LEFT JOIN users AS targetUser ON targetUser.id = alert.targetUserId
->      LEFT JOIN `references` AS reference ON reference.id = alert.referenceId
-> WHERE category.severity = 'info'
-> ORDER BY alert.id DESC
-> LIMIT 40 OFFSET 0;

• | id | select_type |表|类型| possible_keys |关键| key_len | ref |行|额外|
• | 1 |简单|警报|指数| categoryIndex |主要| 8 | NULL | 40 | |
• | 1 |简单|类别| eq_ref | PRIMARY，severityIndex |主要| 8 | am.alert.categoryId | 1 |使用where |
• | 1 |简单|分析仪| eq_ref |主要|主要| 8 | am.alert.analyzerId | 1 | |
• | 1 |简单| analyzerNode | eq_ref |主要|主要| 8 | am.alert.analyzerNodeId | 1 | |
• | 1 |简单| sourceNode | eq_ref |主要|主要| 8 | am.alert.sourceNodeId | 1 | |
• | 1 |简单| targetNode | eq_ref |主要|主要| 8 | am.alert.targetNodeId | 1 | |
• | 1 |简单| sourceUser | eq_ref |主要|主要| 8 | am.alert.sourceUserId | 1 | |
• | 1 |简单| targetUser | eq_ref |主要|主要| 8 | am.alert.targetUserId | 1 | |
• | 1 |简单|参考| eq_ref |主要|主要| 8 | am.alert.referenceId | 1 | |

但是，当我只是更改where子句的值（在这种情况下是枚举，但我不认为相关）时，我可以得到截然不同的结果，我的查询将永远。< / p>

mysql> EXPLAIN SELECT alert.id, category.severity, category.classification, category.completion, alert.description,
->        alert.detectTime, analyzer.name, analyzer.manufacturer, analyzer.model, analyzer.version,
->        analyzer.class, analyzer.osType, analyzer.osVersion, analyzerNode.hostName,
->        analyzerNode.address, sourceNode.hostName, sourceNode.address, targetNode.hostName,
->        targetNode.address, sourceUser.userName, sourceUser.uid, targetUser.userName, targetUser.uid,
->        alert.filePath, alert.sourceProcessPid, alert.sourceProcessPath, alert.targetProcessPid,
->        alert.targetProcessPath, alert.acknowledgeUserName, alert.acknowledgeTime, reference.url, reference.meaning
-> FROM alerts AS alert
->      LEFT JOIN categories AS category ON category.id = alert.categoryId
->      LEFT JOIN analyzers AS analyzer ON analyzer.id = alert.analyzerId
->      LEFT JOIN nodes AS analyzerNode ON analyzerNode.id = alert.analyzerNodeId
->      LEFT JOIN nodes AS sourceNode ON sourceNode.id = alert.sourceNodeId
->      LEFT JOIN nodes AS targetNode ON targetNode.id = alert.targetNodeId
->      LEFT JOIN users AS sourceUser ON sourceUser.id = alert.sourceUserId
->      LEFT JOIN users AS targetUser ON targetUser.id = alert.targetUserId
->      LEFT JOIN `references` AS reference ON reference.id = alert.referenceId
-> WHERE category.severity = 'high'
-> ORDER BY alert.id DESC
-> LIMIT 40 OFFSET 0;

• | id | select_type |表|类型| possible_keys |关键| key_len | ref |行|额外|
• | 1 |简单|类别| ref | PRIMARY，severityIndex | severityIndex | 2 | const | 8 |用在哪里;使用临时;使用filesort |
• | 1 |简单|警报| ref | categoryIndex | categoryIndex | 8 | am.category.id | 3883428 | |
• | 1 |简单|分析仪| eq_ref |主要|主要| 8 | am.alert.analyzerId | 1 | |
• | 1 |简单| analyzerNode | eq_ref |主要|主要| 8 | am.alert.analyzerNodeId | 1 | |
• | 1 |简单| sourceNode | eq_ref |主要|主要| 8 | am.alert.sourceNodeId | 1 | |
• | 1 |简单| targetNode | eq_ref |主要|主要| 8 | am.alert.targetNodeId | 1 | |
• | 1 |简单| sourceUser | eq_ref |主要|主要| 8 | am.alert.sourceUserId | 1 | |
• | 1 |简单| targetUser | eq_ref |主要|主要| 8 | am.alert.targetUserId | 1 | |
• | 1 |简单|参考| eq_ref |主要|主要| 8 | am.alert.referenceId | 1 | |

注意一个人有限制作为执行的第一步，另一个没有。无论如何都要强制MySQL查询优化器首先执行限制吗？

Answer 1

使用ORDER BY和LIMIT时，重要的是订单依据索引。

如果优化程序决定选择另一个索引，则不会应用限制，因为检索结果并扫描更多记录。

在您的情况下，您可能没有“信息”警报更少的“高”，优化程序更喜欢类别表中的严重性索引。

您可以尝试使用FORCE INDEX强制使用警报表的PK，或尝试忽略类别表的严重性索引。

这个article似乎很有用。