为什么MongoDB不在查询中使用复合索引?
问题描述：
以下是此集合的复合索引和单个索引:
Here are the compound index and single index I have for this Collection:
///db.Collection.getIndexes()
/* 1 */
{
"v" : 2,
"key" : {
"_id" : 1
},
"name" : "_id_",
"ns" : "service.Collection"
},
/* 2 */
{
"v" : 2,
"key" : {
"FirstId" : 1,
"SecondId" : 1,
"CreationTime" : -1
},
"name" : "FirstIdSecondIdCreationTime",
"collation" : {
"locale" : "en",
"caseLevel" : false,
"caseFirst" : "off",
"strength" : 1,
"numericOrdering" : false,
"alternate" : "non-ignorable",
"maxVariable" : "punct",
"normalization" : false,
"backwards" : false,
"version" : "57.1"
},
"ns" : "service.Collection"
},
/* 3 */
{
"v" : 2,
"key" : {
"CreationTime" : 1
},
"name" : "CreationTime",
"collation" : {
"locale" : "en",
"caseLevel" : false,
"caseFirst" : "off",
"strength" : 1,
"numericOrdering" : false,
"alternate" : "non-ignorable",
"maxVariable" : "punct",
"normalization" : false,
"backwards" : false,
"version" : "57.1"
},
"ns" : "service.Collection"
}
预期结果是使用 FirstIdSecondIdCreationTime 索引的IXSCAN:
The expected result is an IXSCAN using the FirstIdSecondIdCreationTime index:
///service.Collection.find({ FirstId: "771367b7-4bef-49ab-bda1-6230254c6349", ///SecondId: "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f" })
/// .projection({})
/// .sort({_id:-1}).hint("FirstIdSecondIdCreationTime").explain('executionStats')
{
"queryPlanner" : {
"plannerVersion" : 1,
"namespace" : "service.Collection",
"indexFilterSet" : false,
"parsedQuery" : {
"$and" : [
{
"FirstId" : {
"$eq" : "771367b7-4bef-49ab-bda1-6230254c6349"
}
},
{
"SecondId" : {
"$eq" : "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f"
}
}
]
},
"winningPlan" : {
"stage" : "SORT",
"sortPattern" : {
"_id" : -1
},
"inputStage" : {
"stage" : "SORT_KEY_GENERATOR",
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"FirstId" : {
"$eq" : "771367b7-4bef-49ab-bda1-6230254c6349"
}
},
{
"SecondId" : {
"$eq" : "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f"
}
}
]
},
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : {
"FirstId" : 1,
"SecondId" : 1,
"CreationTime" : -1
},
"indexName" : "FirstIdSecondIdCreationTime",
"collation" : {
"locale" : "en",
"caseLevel" : false,
"caseFirst" : "off",
"strength" : 1,
"numericOrdering" : false,
"alternate" : "non-ignorable",
"maxVariable" : "punct",
"normalization" : false,
"backwards" : false,
"version" : "57.1"
},
"isMultiKey" : false,
"multiKeyPaths" : {
"FirstId" : [ ],
"SecondId" : [ ],
"CreationTime" : [ ]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"FirstId" : [
"[MinKey, MaxKey]"
],
"SecondId" : [
"[MinKey, MaxKey]"
],
"CreationTime" : [
"[MaxKey, MinKey]"
]
}
}
}
}
},
"rejectedPlans" : [ ]
},
"executionStats" : {
"executionSuccess" : true,
"nReturned" : 1,
"executionTimeMillis" : 5491,
"totalKeysExamined" : 856730,
"totalDocsExamined" : 856730,
"executionStages" : {
"stage" : "SORT",
"nReturned" : 1,
"executionTimeMillisEstimate" : 5261,
"works" : 856734,
"advanced" : 1,
"needTime" : 856732,
"needYield" : 0,
"saveState" : 6697,
"restoreState" : 6697,
"isEOF" : 1,
"invalidates" : 0,
"sortPattern" : {
"_id" : -1
},
"memUsage" : 432,
"memLimit" : 33554432,
"inputStage" : {
"stage" : "SORT_KEY_GENERATOR",
"nReturned" : 1,
"executionTimeMillisEstimate" : 5201,
"works" : 856732,
"advanced" : 1,
"needTime" : 856730,
"needYield" : 0,
"saveState" : 6697,
"restoreState" : 6697,
"isEOF" : 1,
"invalidates" : 0,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"FirstId" : {
"$eq" : "771367b7-4bef-49ab-bda1-6230254c6349"
}
},
{
"SecondId" : {
"$eq" : "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f"
}
}
]
},
"nReturned" : 1,
"executionTimeMillisEstimate" : 5131,
"works" : 856731,
"advanced" : 1,
"needTime" : 856729,
"needYield" : 0,
"saveState" : 6697,
"restoreState" : 6697,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 856730,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 856730,
"executionTimeMillisEstimate" : 820,
"works" : 856731,
"advanced" : 856730,
"needTime" : 0,
"needYield" : 0,
"saveState" : 6697,
"restoreState" : 6697,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"FirstId" : 1,
"SecondId" : 1,
"CreationTime" : -1
},
"indexName" : "FirstIdSecondIdCreationTime",
"collation" : {
"locale" : "en",
"caseLevel" : false,
"caseFirst" : "off",
"strength" : 1,
"numericOrdering" : false,
"alternate" : "non-ignorable",
"maxVariable" : "punct",
"normalization" : false,
"backwards" : false,
"version" : "57.1"
},
"isMultiKey" : false,
"multiKeyPaths" : {
"FirstId" : [ ],
"SecondId" : [ ],
"CreationTime" : [ ]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"FirstId" : [
"[MinKey, MaxKey]"
],
"SecondId" : [
"[MinKey, MaxKey]"
],
"CreationTime" : [
"[MaxKey, MinKey]"
]
},
"keysExamined" : 856730,
"seeks" : 1,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0,
"indexDef" : {
"indexName" : "FirstIdSecondIdCreationTime",
"isMultiKey" : false,
"multiKeyPaths" : {
"FirstId" : [ ],
"SecondId" : [ ],
"CreationTime" : [ ]
},
"keyPattern" : {
"FirstId" : 1,
"SecondId" : 1,
"CreationTime" : -1
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"direction" : "forward"
}
}
}
}
}
但实际结果是COLLSCAN占用了8000毫秒以上的时间:
but the actual result is a COLLSCAN that takes over 8000ms:
"event": {
"dataset": "mongodb.log",
"module": "mongodb"
},
"service": {
"type": "mongodb"
},
"message": "command service.Collection command: find { find: \"Collection\",
filter: { FirstId: \"771367b7-4bef-49ab-bda1-6230254c6349\", SecondId: \"3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f\" }, sort: { CreationTime: -1 }, limit: 1,
planSummary: COLLSCAN keysExamined:0 docsExamined:784787 hasSortStage:1 cursorExhausted:1 numYields:6175 nreturned:1 reslen:677
locks:{ Global: { acquireCount: { r: 12352 } }, Database: { acquireCount: { r: 6176 } }, Collection: { acquireCount: { r: 6176 } } } protocol:op_msg 8441ms",
"mongodb.docsExamined": 784787,
"fileset": {
"name": "log"
},
为什么使用 FirstIdSecondIDCreationTime 复合索引进行COLLSCANing而不是IXSCANing?有没有一种方法可以更改我的索引/查询来加快查询速度?
Why am I COLLSCANing instead of IXSCANing with the FirstIdSecondIDCreationTime compound index? Is there a way to change my index/ my query to speed up the query?
根据评论中的建议，我运行了 explain("allPlansExecution").
Per a suggestion in the comments, I've run explain("allPlansExecution").
///db.Collection.find({ FirstId: "771367b7-4bef-49ab-bda1-6230254c6349", ///SecondId: "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f" })
/// .projection({})
/// .sort({_id:-1}).explain('allPlansExecution')
{
"queryPlanner" : {
"plannerVersion" : 1,
"namespace" : "service.Collection",
"indexFilterSet" : false,
"parsedQuery" : {
"$and" : [
{
"FirstId" : {
"$eq" : "771367b7-4bef-49ab-bda1-6230254c6349"
}
},
{
"SecondId" : {
"$eq" : "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f"
}
}
]
},
"winningPlan" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"FirstId" : {
"$eq" : "771367b7-4bef-49ab-bda1-6230254c6349"
}
},
{
"SecondId" : {
"$eq" : "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f"
}
}
]
},
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : {
"_id" : 1
},
"indexName" : "_id_",
"isMultiKey" : false,
"multiKeyPaths" : {
"_id" : [ ]
},
"isUnique" : true,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "backward",
"indexBounds" : {
"_id" : [
"[MaxKey, MinKey]"
]
}
}
},
"rejectedPlans" : [ ]
},
"executionStats" : {
"executionSuccess" : true,
"nReturned" : 1,
"executionTimeMillis" : 5408,
"totalKeysExamined" : 856748,
"totalDocsExamined" : 856748,
"executionStages" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"FirstId" : {
"$eq" : "771367b7-4bef-49ab-bda1-6230254c6349"
}
},
{
"SecondId" : {
"$eq" : "3bffb3cd-fb5e-43e5-abd1-e0b48c97f78f"
}
}
]
},
"nReturned" : 1,
"executionTimeMillisEstimate" : 4862,
"works" : 856749,
"advanced" : 1,
"needTime" : 856747,
"needYield" : 0,
"saveState" : 6694,
"restoreState" : 6694,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 856748,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 856748,
"executionTimeMillisEstimate" : 1220,
"works" : 856749,
"advanced" : 856748,
"needTime" : 0,
"needYield" : 0,
"saveState" : 6694,
"restoreState" : 6694,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"_id" : 1
},
"indexName" : "_id_",
"isMultiKey" : false,
"multiKeyPaths" : {
"_id" : [ ]
},
"isUnique" : true,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "backward",
"indexBounds" : {
"_id" : [
"[MaxKey, MinKey]"
]
},
"keysExamined" : 856748,
"seeks" : 1,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
},
"allPlansExecution" : [ ]
}
}
答
"FirstIdSecondIdCreationTime"索引没有自动考虑，因为它是使用排序规则创建的，并且查询是在没有排序规则的情况下运行的.
The "FirstIdSecondIdCreationTime" index was not automatically considered because it was created with a collation, and the query is being run without a collation.
使用 .collation()游标方法，可为用于索引的查询指定相同的排序规则.
Use the .collation() cursor method to specify the same collation for the query that was used for the index.
使用该索引的5.5秒运行时间也相当慢.如果您在 {FirstId:1，SecondId:1，_id:1} 上创建索引，则该查询可能会有所改进，以便他们查询执行程序可以使用索引来满足排序要求，而不是内存中排序.
The 5.5 second run time using that index is pretty slow as well.
You may see some improvement in that query if you create an index on {FirstId: 1, SecondId: 1, _id: 1} so that they query executor can use the index to meet the sort instead of an in-memory sort.