即使有索引，MongoDB 也发现查询非常慢
Posted
技术标签:
【中文标题】即使有索引，MongoDB 也发现查询非常慢【英文标题】：MongoDB find query very slow even with index 【发布时间】：2019-03-01 02:44:28 【问题描述】：我不明白为什么下面的查询这么慢，解释显示它需要 74 多秒才能完成，尽管存在索引。
dev_id 和 _id 都已编入索引，我可以向您保证，这似乎根本没有帮助。
db.DeviceLoginLog.find("dev_id": "xxx").skip(0).limit(10).sort("_id": -1).explain("executionStats");
谁能帮我解释解释输出并建议我如何加快速度？
"queryPlanner" :
"plannerVersion" : 1,
"namespace" : "example.DeviceLoginLog",
"indexFilterSet" : false,
"parsedQuery" :
"dev_id" :
"$eq" : "xxx"
,
"winningPlan" :
"stage" : "SORT",
"sortPattern" :
"_id" : -1
,
"limitAmount" : 10,
"inputStage" :
"stage" : "KEEP_MUTATIONS",
"inputStage" :
"stage" : "FETCH",
"inputStage" :
"stage" : "IXSCAN",
"keyPattern" :
"dev_id" : 1
,
"indexName" : "dev_id_1",
"isMultiKey" : false,
"direction" : "forward",
"indexBounds" :
"dev_id" : [
"[\"xxx\", \"xxx\"]"
]
,
"rejectedPlans" : [
"stage" : "LIMIT",
"limitAmount" : 4,
"inputStage" :
"stage" : "FETCH",
"filter" :
"dev_id" :
"$eq" : "xxx"
,
"inputStage" :
"stage" : "IXSCAN",
"keyPattern" :
"_id" : 1
,
"indexName" : "_id_",
"isMultiKey" : false,
"direction" : "backward",
"indexBounds" :
"_id" : [
"[MaxKey, MinKey]"
]
]
,
"executionStats" :
"executionSuccess" : true,
"nReturned" : 10,
"executionTimeMillis" : 74867,
"totalKeysExamined" : 9203,
"totalDocsExamined" : 9203,
"executionStages" :
"stage" : "SORT",
"nReturned" : 10,
"executionTimeMillisEstimate" : 49860,
"works" : 9537,
"advanced" : 10,
"needTime" : 9204,
"needFetch" : 321,
"saveState" : 1864,
"restoreState" : 1864,
"isEOF" : 1,
"invalidates" : 1064,
"sortPattern" :
"_id" : -1
,
"memUsage" : 1890,
"memLimit" : 33554432,
"limitAmount" : 10,
"inputStage" :
"stage" : "KEEP_MUTATIONS",
"nReturned" : 9203,
"executionTimeMillisEstimate" : 49820,
"works" : 9525,
"advanced" : 9203,
"needTime" : 0,
"needFetch" : 321,
"saveState" : 1864,
"restoreState" : 1864,
"isEOF" : 1,
"invalidates" : 1064,
"inputStage" :
"stage" : "FETCH",
"nReturned" : 9203,
"executionTimeMillisEstimate" : 49820,
"works" : 9525,
"advanced" : 9203,
"needTime" : 0,
"needFetch" : 321,
"saveState" : 1864,
"restoreState" : 1864,
"isEOF" : 1,
"invalidates" : 1064,
"docsExamined" : 9203,
"alreadyHasObj" : 0,
"inputStage" :
"stage" : "IXSCAN",
"nReturned" : 9203,
"executionTimeMillisEstimate" : 10,
"works" : 9204,
"advanced" : 9203,
"needTime" : 0,
"needFetch" : 0,
"saveState" : 1864,
"restoreState" : 1864,
"isEOF" : 1,
"invalidates" : 1064,
"keyPattern" :
"dev_id" : 1
,
"indexName" : "dev_id_1",
"isMultiKey" : false,
"direction" : "forward",
"indexBounds" :
"dev_id" : [
"[\"xxx\", \"xxx\"]"
]
,
"keysExamined" : 9203,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0,
"matchTested" : 0
,
"serverInfo" :
"host" : "iZ231ear7c9Z",
"port" : 27017,
"version" : "3.0.3",
"gitVersion" : "b40106b36eecd1b4407eb1ad1af6bc60593c6105"
,
"ok" : 1
提前致谢！
【问题讨论】：
【参考方案1】：查询使用索引dev_id: 1 进行查找，然后必须将匹配的文档提取到内存中以便在不使用索引的情况下进行排序。创建一个复合索引为dev_id: 1, _id: -1 将提高性能。
见sort-and-non-prefix-subset-of-an-index。
【讨论】：
【参考方案2】："totalKeysExamined" : 9203,
"totalDocsExamined" : 9203,
这表明查询必须检查每个单独的文档。您没有在 _id 字段上使用索引。此外，根据您要返回的字段，考虑创建一个允许覆盖查询的索引。
覆盖查询是可以完全使用 索引并且不必检查任何文档。一个索引涵盖一个 当以下两个都适用时查询：
查询中的所有字段都是索引的一部分，并且 结果中返回的所有字段都在同一个索引中。
https://docs.mongodb.com/manual/core/query-optimization/
【讨论】：
不适用于我的情况，我需要返回整个文档。以上是关于即使有索引，MongoDB 也发现查询非常慢的主要内容，如果未能解决你的问题，请参考以下文章