I'm going through the existing indexes on a fairly old MongoDB database which has recently been upgraded from 2.something to 3.4.
The code accessing it is convoluted, automagic, and spaghetti-heavy, plus it allows web-facing frontend code to do arbitrary queries, so it's pretty difficult to get an overview of exactly what queries are made. Doing a bit of profiling, I quickly discovered that a lot of queries were missing indexes, either because logical indexes were missing or misspelled (yeah!), or because someone had assumed that a compound index would support queries on non-prefix parts of them.
Such an example was with a users collection where a compound index, { boxes._id: 1, _deleted: 1 } was missed by a ton of queries filtering by { _deleted: { $exists: false } }.
I wiped it, and created single-field indexes instead, having read up on the index intersection feature which is now part of the version of the MongoDB server backing this app:
db.users.dropIndex({ 'boxes._id': 1, _deleted: 1 });
db.users.ensureIndex({ 'boxes._id': 1 });
db.users.ensureIndex({ '_deleted': 1 });
So far, so good. The dataset here is not very large, so I saw no immediate issue, performance-wise, by dropping the compound index and support index intersection instead. AFAIU from the docs and the interwebs, the performance gains from using compound indexes are mostly felt with really large collections, and for this one, we're looking at about 11k documents.
Simple count queries on either of the two indexed fields yield satisfying responses while hitting the keys:
db.runCommand({ explain: { count: 'users', query: { 'boxes._id': ObjectId('597745846ca2582d8b364c38') }, verbosity: 'executionStats' }})
// "executionTimeMillis" : 8
db.runCommand({ explain: { count: 'users', query: { _deleted: { $exists: false } }, verbosity: 'executionStats' }})
// "executionTimeMillis" : 35
(Note that this is on my old MacBook Air, and compared to query times I'm seeing in the performance log when not using explain, then "normal" response times seem to be less than 25% of those reported with explain in the mongo shell)
However, if I filter on both those fields thus triggering the use of an intersection index, I see a massive performance penalty:
{
"queryPlanner" : {
"plannerVersion" : 1,
"namespace" : "bl.users",
"indexFilterSet" : false,
"parsedQuery" : {
"$and" : [
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"winningPlan" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
}
]
},
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : { "_deleted" : 1 },
"indexName" : "_deleted_1",
"isMultiKey" : false,
"multiKeyPaths" : { "_deleted" : [ ] },
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : { "_deleted" : [ "[null, null]" ] }
}
},
"rejectedPlans" : [
{
"stage" : "FETCH",
"filter" : {
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : {
"boxes._id" : 1
},
"indexName" : "boxes._id_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
]
}
}
},
{
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"inputStage" : {
"stage" : "AND_SORTED",
"inputStages" : [
{
"stage" : "IXSCAN",
"keyPattern" : {
"boxes._id" : 1
},
"indexName" : "boxes._id_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
]
}
},
{
"stage" : "IXSCAN",
"keyPattern" : { "_deleted" : 1 },
"indexName" : "_deleted_1",
"isMultiKey" : false,
"multiKeyPaths" : { "_deleted" : [ ] },
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"_deleted" : [ "[null, null]" ]
}
}
]
}
}
]
},
"executionStats" : {
"executionSuccess" : true,
"nReturned" : 11098,
"executionTimeMillis" : 731,
"totalKeysExamined" : 11098,
"totalDocsExamined" : 11098,
"executionStages" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
}
]
},
"nReturned" : 11098,
"executionTimeMillisEstimate" : 702,
"works" : 11099,
"advanced" : 11098,
"needTime" : 0,
"needYield" : 0,
"saveState" : 127,
"restoreState" : 127,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 11098,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 11098,
"executionTimeMillisEstimate" : 49,
"works" : 11099,
"advanced" : 11098,
"needTime" : 0,
"needYield" : 0,
"saveState" : 127,
"restoreState" : 127,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : { "_deleted" : 1 },
"indexName" : "_deleted_1",
"isMultiKey" : false,
"multiKeyPaths" : { "_deleted" : [ ] },
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"_deleted" : [ "[null, null]" ]
},
"keysExamined" : 11098,
"seeks" : 1,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
},
"allPlansExecution" : [
{
"nReturned" : 101,
"executionTimeMillisEstimate" : 0,
"totalKeysExamined" : 101,
"totalDocsExamined" : 101,
"executionStages" : {
"stage" : "FETCH",
"filter" : {
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
"nReturned" : 101,
"executionTimeMillisEstimate" : 0,
"works" : 101,
"advanced" : 101,
"needTime" : 0,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"docsExamined" : 101,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 101,
"executionTimeMillisEstimate" : 0,
"works" : 101,
"advanced" : 101,
"needTime" : 0,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"keyPattern" : {
"boxes._id" : 1
},
"indexName" : "boxes._id_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
]
},
"keysExamined" : 101,
"seeks" : 1,
"dupsTested" : 101,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
},
{
"nReturned" : 50,
"executionTimeMillisEstimate" : 12,
"totalKeysExamined" : 101,
"totalDocsExamined" : 50,
"executionStages" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"nReturned" : 50,
"executionTimeMillisEstimate" : 12,
"works" : 101,
"advanced" : 50,
"needTime" : 51,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"docsExamined" : 50,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "AND_SORTED",
"nReturned" : 50,
"executionTimeMillisEstimate" : 12,
"works" : 101,
"advanced" : 50,
"needTime" : 51,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"flagged" : 0,
"failedAnd_0" : 0,
"failedAnd_1" : 0,
"inputStages" : [
{
"stage" : "IXSCAN",
"nReturned" : 51,
"executionTimeMillisEstimate" : 0,
"works" : 51,
"advanced" : 51,
"needTime" : 0,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"keyPattern" : {
"boxes._id" : 1
},
"indexName" : "boxes._id_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
]
},
"keysExamined" : 51,
"seeks" : 1,
"dupsTested" : 51,
"dupsDropped" : 0,
"seenInvalidated" : 0
},
{
"stage" : "IXSCAN",
"nReturned" : 50,
"executionTimeMillisEstimate" : 12,
"works" : 50,
"advanced" : 50,
"needTime" : 0,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"keyPattern" : { "_deleted" : 1 },
"indexName" : "_deleted_1",
"isMultiKey" : false,
"multiKeyPaths" : { "_deleted" : [ ] },
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"_deleted" : [ "[null, null]" ]
},
"keysExamined" : 50,
"seeks" : 1,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
]
}
}
},
{
"nReturned" : 101,
"executionTimeMillisEstimate" : 11,
"totalKeysExamined" : 101,
"totalDocsExamined" : 101,
"executionStages" : {
"stage" : "FETCH",
"filter" : {
"$and" : [
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
}
]
},
"nReturned" : 101,
"executionTimeMillisEstimate" : 11,
"works" : 101,
"advanced" : 101,
"needTime" : 0,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"docsExamined" : 101,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 101,
"executionTimeMillisEstimate" : 0,
"works" : 101,
"advanced" : 101,
"needTime" : 0,
"needYield" : 0,
"saveState" : 3,
"restoreState" : 3,
"isEOF" : 0,
"invalidates" : 0,
"keyPattern" : { "_deleted" : 1 },
"indexName" : "_deleted_1",
"isMultiKey" : false,
"multiKeyPaths" : { "_deleted" : [ ] },
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"_deleted" : [ "[null, null]" ]
},
"keysExamined" : 101,
"seeks" : 1,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
}
]
}
}
The same query hitting the compound index:
{
"queryPlanner" : {
"plannerVersion" : 1,
"namespace" : "bl.users",
"indexFilterSet" : false,
"parsedQuery" : {
"$and" : [
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"winningPlan" : {
"stage" : "COUNT",
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : {
"boxes._id" : 1,
"_deleted" : 1
},
"indexName" : "boxes._id_1__deleted_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
],
"_deleted" : [ ]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
],
"_deleted" : [ "[null, null]" ]
}
}
}
},
"rejectedPlans" : [ ]
},
"executionStats" : {
"executionSuccess" : true,
"nReturned" : 0,
"executionTimeMillis" : 37,
"totalKeysExamined" : 11098,
"totalDocsExamined" : 11098,
"executionStages" : {
"stage" : "COUNT",
"nReturned" : 0,
"executionTimeMillisEstimate" : 35,
"works" : 11099,
"advanced" : 0,
"needTime" : 11098,
"needYield" : 0,
"saveState" : 87,
"restoreState" : 87,
"isEOF" : 1,
"invalidates" : 0,
"nCounted" : 11098,
"nSkipped" : 0,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
"nReturned" : 11098,
"executionTimeMillisEstimate" : 35,
"works" : 11099,
"advanced" : 11098,
"needTime" : 0,
"needYield" : 0,
"saveState" : 87,
"restoreState" : 87,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 11098,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 11098,
"executionTimeMillisEstimate" : 23,
"works" : 11099,
"advanced" : 11098,
"needTime" : 0,
"needYield" : 0,
"saveState" : 87,
"restoreState" : 87,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"boxes._id" : 1,
"_deleted" : 1
},
"indexName" : "boxes._id_1__deleted_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
],
"_deleted" : [ ]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
],
"_deleted" : [ "[null, null]" ]
},
"keysExamined" : 11098,
"seeks" : 1,
"dupsTested" : 11098,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
},
"allPlansExecution" : [ ]
}
}
Replacing the compound index with just a single index on boxes._id yields pretty much identical performance to having the compound key on both filtered fields:
{
"queryPlanner" : {
"plannerVersion" : 1,
"namespace" : "bl.users",
"indexFilterSet" : false,
"parsedQuery" : {
"$and" : [
{
"boxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"winningPlan" : {
"stage" : "COUNT",
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : {
"boxes._id" : 1
},
"indexName" : "boxes._id_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
]
}
}
}
},
"rejectedPlans" : [ ]
},
"executionStats" : {
"executionSuccess" : true,
"nReturned" : 0,
"executionTimeMillis" : 31,
"totalKeysExamined" : 11098,
"totalDocsExamined" : 11098,
"executionStages" : {
"stage" : "COUNT",
"nReturned" : 0,
"executionTimeMillisEstimate" : 34,
"works" : 11099,
"advanced" : 0,
"needTime" : 11098,
"needYield" : 0,
"saveState" : 88,
"restoreState" : 88,
"isEOF" : 1,
"invalidates" : 0,
"nCounted" : 11098,
"nSkipped" : 0,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
},
"nReturned" : 11098,
"executionTimeMillisEstimate" : 34,
"works" : 11099,
"advanced" : 11098,
"needTime" : 0,
"needYield" : 0,
"saveState" : 88,
"restoreState" : 88,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 11098,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 11098,
"executionTimeMillisEstimate" : 11,
"works" : 11099,
"advanced" : 11098,
"needTime" : 0,
"needYield" : 0,
"saveState" : 88,
"restoreState" : 88,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"boxes._id" : 1
},
"indexName" : "boxes._id_1",
"isMultiKey" : true,
"multiKeyPaths" : {
"boxes._id" : [
"boxes"
]
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"boxes._id" : [
"[ObjectId('597745846ca2582d8b364c38'), ObjectId('597745846ca2582d8b364c38')]"
]
},
"keysExamined" : 11098,
"seeks" : 1,
"dupsTested" : 11098,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
},
"allPlansExecution" : [ ]
}
}
Finally, dropping all indexes on these fields, triggering a COLLSCAN yields performance pretty much identical to the index intersection one:
{
"queryPlanner" : {
"plannerVersion" : 1,
"namespace" : "bl.users",
"indexFilterSet" : false,
"parsedQuery" : {
"$and" : [
{
"ideaboxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"winningPlan" : {
"stage" : "COUNT",
"inputStage" : {
"stage" : "COLLSCAN",
"filter" : {
"$and" : [
{
"ideaboxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"direction" : "forward"
}
},
"rejectedPlans" : [ ]
},
"executionStats" : {
"executionSuccess" : true,
"nReturned" : 0,
"executionTimeMillis" : 848,
"totalKeysExamined" : 0,
"totalDocsExamined" : 11109,
"executionStages" : {
"stage" : "COUNT",
"nReturned" : 0,
"executionTimeMillisEstimate" : 847,
"works" : 11111,
"advanced" : 0,
"needTime" : 11110,
"needYield" : 0,
"saveState" : 123,
"restoreState" : 123,
"isEOF" : 1,
"invalidates" : 0,
"nCounted" : 11098,
"nSkipped" : 0,
"inputStage" : {
"stage" : "COLLSCAN",
"filter" : {
"$and" : [
{
"ideaboxes._id" : {
"$eq" : ObjectId("597745846ca2582d8b364c38")
}
},
{
"$nor" : [ { "_deleted" : { "$exists" : true } } ]
}
]
},
"nReturned" : 11098,
"executionTimeMillisEstimate" : 847,
"works" : 11111,
"advanced" : 11098,
"needTime" : 12,
"needYield" : 0,
"saveState" : 123,
"restoreState" : 123,
"isEOF" : 1,
"invalidates" : 0,
"direction" : "forward",
"docsExamined" : 11109
}
},
"allPlansExecution" : [ ]
}
}
Surely getting COLLSCAN performance with index intersection queries can't be right? I admit I'm very new with Mongo, so I'm also quite ignorant when trying to decipher this report — the answer may be staring me in the face.
I've tested the same queries on Mongo 3.2 on an ObjectRocket instance, and while their servers are much faster than my ageing laptop, they exhibit the same numbers relatively speaking.
Any ideas why this is happening?
Thank you :)
