How to improve mongo queries with multikey index in array - mongodb

Question

I just found out that 1 of our queries is quite slow. We use a multikey index in an array and not sure how we can actually improve this query.
Our data structure:
account: [{
accountId:string,
service:string
}]
Index key is:
{"account.accountId":1, "account.service":1}
Here's my query:
db.model.find({"account.accountId":"12345", "account.service":"GameCenter"}).explain();
executionStats (1 million records)
"executionStats" : {
"nReturned" : 1,
"executionTimeMillis" : 325,
"totalKeysExamined" : 122744,
"totalDocsExamined" : 122743,
"executionStages" : {
"stage" : "SHARD_MERGE",
"nReturned" : 1,
"executionTimeMillis" : 325,
"totalKeysExamined" : 122744,
"totalDocsExamined" : 122743,
"totalChildMillis" : NumberLong(834),
"shards" : [
{
"shardName" : "rs1",
"executionSuccess" : true,
"executionStages" : {
"stage" : "SHARDING_FILTER",
"nReturned" : 0,
"executionTimeMillisEstimate" : 300,
"works" : 40999,
"advanced" : 0,
"needTime" : 40998,
"needYield" : 0,
"saveState" : 320,
"restoreState" : 320,
"isEOF" : 1,
"invalidates" : 0,
"chunkSkips" : 0,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"account.accountId" : {
"$eq" : "G:8183971619"
}
},
"nReturned" : 0,
"executionTimeMillisEstimate" : 300,
"works" : 40999,
"advanced" : 0,
"needTime" : 40998,
"needYield" : 0,
"saveState" : 320,
"restoreState" : 320,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 40998,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 40998,
"executionTimeMillisEstimate" : 30,
"works" : 40999,
"advanced" : 40998,
"needTime" : 0,
"needYield" : 0,
"saveState" : 320,
"restoreState" : 320,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"account.service" : 1,
"account.accountId" : 1
},
"indexName" : "account.service_1_account.accountId_1",
"isMultiKey" : true,
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 1,
"direction" : "forward",
"indexBounds" : {
"account.service" : [
"[\"GameCenter\", \"GameCenter\"]"
],
"account.accountId" : [
"[MinKey, MaxKey]"
]
},
"keysExamined" : 40998,
"dupsTested" : 40998,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
}
},
{
"shardName" : "rs2",
"executionSuccess" : true,
"executionStages" : {
"stage" : "SHARDING_FILTER",
"nReturned" : 1,
"executionTimeMillisEstimate" : 240,
"works" : 40612,
"advanced" : 1,
"needTime" : 40610,
"needYield" : 0,
"saveState" : 317,
"restoreState" : 317,
"isEOF" : 1,
"invalidates" : 0,
"chunkSkips" : 0,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"account.accountId" : {
"$eq" : "G:8183971619"
}
},
"nReturned" : 1,
"executionTimeMillisEstimate" : 230,
"works" : 40612,
"advanced" : 1,
"needTime" : 40610,
"needYield" : 0,
"saveState" : 317,
"restoreState" : 317,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 40610,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 40611,
"executionTimeMillisEstimate" : 70,
"works" : 40612,
"advanced" : 40611,
"needTime" : 0,
"needYield" : 0,
"saveState" : 317,
"restoreState" : 317,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"account.service" : 1,
"account.accountId" : 1
},
"indexName" : "account.service_1_account.accountId_1",
"isMultiKey" : true,
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 1,
"direction" : "forward",
"indexBounds" : {
"account.service" : [
"[\"GameCenter\", \"GameCenter\"]"
],
"account.accountId" : [
"[MinKey, MaxKey]"
]
},
"keysExamined" : 40611,
"dupsTested" : 40611,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
}
},
{
"shardName" : "rs3",
"executionSuccess" : true,
"executionStages" : {
"stage" : "SHARDING_FILTER",
"nReturned" : 0,
"executionTimeMillisEstimate" : 270,
"works" : 41136,
"advanced" : 0,
"needTime" : 41135,
"needYield" : 0,
"saveState" : 321,
"restoreState" : 321,
"isEOF" : 1,
"invalidates" : 0,
"chunkSkips" : 0,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"account.accountId" : {
"$eq" : "G:8183971619"
}
},
"nReturned" : 0,
"executionTimeMillisEstimate" : 270,
"works" : 41136,
"advanced" : 0,
"needTime" : 41135,
"needYield" : 0,
"saveState" : 321,
"restoreState" : 321,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 41135,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 41135,
"executionTimeMillisEstimate" : 90,
"works" : 41136,
"advanced" : 41135,
"needTime" : 0,
"needYield" : 0,
"saveState" : 321,
"restoreState" : 321,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"account.service" : 1,
"account.accountId" : 1
},
"indexName" : "account.service_1_account.accountId_1",
"isMultiKey" : true,
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 1,
"direction" : "forward",
"indexBounds" : {
"account.service" : [
"[\"GameCenter\", \"GameCenter\"]"
],
"account.accountId" : [
"[MinKey, MaxKey]"
]
},
"keysExamined" : 41135,
"dupsTested" : 41135,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
}
}
]
}
},
To summarise, the query do use the indexes but scan a lot of documents to get the results (122k / 1million). We couldn't use covered index as we need to return some other information too.
Also something werid: I'm querying for a unique accoundId, but my indexBounds for account.accountId is:
"account.accountId" : [
"[MinKey, MaxKey]"
]
Thank you for your help!
Mars

Answer 1

After reading the executionStatus carefully, the query is actually scanning all indexes since "accountId" indexBounds is [MinKey, MaxKey].
I'm able to resolve this issue with $elemMatch
db.model.find({
account:{
$elemMatch:{
accountId:"111",
service:"facebook"
}
}
})