MongoDB 数组的性能问题

Question

我正在寻找有关如何提高查询性能的建议。

我在 mongoose 中有这个用户模型，我正在索引 interested_cities.

const firebaseToken = new Schema({
    type: {
        type: String,
        default: "android",
        required: true,
        trim: true,
    },
    device_id: {
        type: String,
        required: true,
        trim: true,
    },
    fcm_token: {
        type: String,
        required: true,
        trim: true,
    },
});

const userSchema = new Schema({
    name: {
        type: String,
        required: true,
        trim: true,
    },
    interested_cities: {
        type: [{
            type: String,
            trim: true,
            lowercase: true,
            unique: true
        }],
        required: false,
        default: [],
    },
    push_notification_tokens_firebase: {
        type: [firebaseToken],
        required: false,
        default: [],
    },
});

userSchema.index({
    interested_cities: 1
});

我正在寻找的是查询在 interested_cities 数组中具有 'A' 或 'B' 的用户。

我正在查询类似的内容。我只需要查询中的 firebase fcm_token。

const involvedUsers = await User.find(
    {
      $or: [
        { interested_cities: { $in: ['A', 'B'] } },
        { phone_number: { $in: adminPhoneNumbersList } },
      ],
    },
    {
      _id: 1,
      "push_notification_tokens_firebase.fcm_token": 1,
    }
  );

目前，查询 14k 文档需要 20 秒，这需要改进。任何指针将不胜感激。

解释：

{
                "explainVersion": "1",
                "queryPlanner": {
                    "namespace": "production.users",
                    "indexFilterSet": false,
                    "parsedQuery": {
                        "interested_cities": {
                            "$in": [
                                "A",
                                "B"
                            ]
                        }
                    },
                    "maxIndexedOrSolutionsReached": false,
                    "maxIndexedAndSolutionsReached": false,
                    "maxScansToExplodeReached": false,
                    "winningPlan": {
                        "stage": "PROJECTION_DEFAULT",
                        "transformBy": {
                            "_id": 1,
                            "push_notification_tokens_firebase.fcm_token": 1
                        },
                        "inputStage": {
                            "stage": "FETCH",
                            "inputStage": {
                                "stage": "IXSCAN",
                                "keyPattern": {
                                    "interested_cities": 1
                                },
                                "indexName": "interested_cities_1",
                                "isMultiKey": true,
                                "multiKeyPaths": {
                                    "interested_cities": [
                                        "interested_cities"
                                    ]
                                },
                                "isUnique": false,
                                "isSparse": false,
                                "isPartial": false,
                                "indexVersion": 2,
                                "direction": "forward",
                                "indexBounds": {
                                    "interested_cities": [
                                        "[\"A\", \"A\"]",
                                        "[\"B\", \"B\"]"
                                    ]
                                }
                            }
                        }
                    },
                    "rejectedPlans": []
                }

                "executionStats": {
                    "executionSuccess": true,
                    "nReturned": 6497,
                    "executionTimeMillis": 48,
                    "totalKeysExamined": 0,
                    "totalDocsExamined": 14827,
                    "executionStages": {
                        "stage": "SUBPLAN",
                        "nReturned": 6497,
                        "executionTimeMillisEstimate": 46,
                        "works": 14829,
                        "advanced": 6497,
                        "needTime": 8331,
                        "needYield": 0,
                        "saveState": 14,
                        "restoreState": 14,
                        "isEOF": 1,
                        "inputStage": {
                            "stage": "PROJECTION_DEFAULT",
                            "nReturned": 6497,
                            "executionTimeMillisEstimate": 46,
                            "works": 14829,
                            "advanced": 6497,
                            "needTime": 8331,
                            "needYield": 0,
                            "saveState": 14,
                            "restoreState": 14,
                            "isEOF": 1,
                            "transformBy": {
                                "_id": 1,
                                "push_notification_tokens_firebase.fcm_token": 1
                            },
                            "inputStage": {
                                "stage": "COLLSCAN",
                                "filter": {
                                    "$or": [
                                        {
                                            "interested_cities": {
                                                "$in": [
                                                    "A",
                                                    "B"
                                                ]
                                            }
                                        },
                                        {
                                            "phone_number": {
                                                "$in": [
                                                    "phone numbers",
                                                    "phone number"
                                                ]
                                            }
                                        }
                                    ]
                                },
                                "nReturned": 6497,
                                "executionTimeMillisEstimate": 41,
                                "works": 14829,
                                "advanced": 6497,
                                "needTime": 8331,
                                "needYield": 0,
                                "saveState": 14,
                                "restoreState": 14,
                                "isEOF": 1,
                                "direction": "forward",
                                "docsExamined": 14827
                            }
                        }
                    },
                    "allPlansExecution": []
                }

Answer 1

猫鼬优化：

By default, Mongoose queries return an instance of the Mongoose Document class. Documents are much heavier than vanilla JavaScript objects, because they have a lot of internal state for change tracking. Enabling the lean option tells Mongoose to skip instantiating a full Mongoose document and just give you the POJO.

https://mongoosejs.com/docs/tutorials/lean.html#using-lean

您可以在 per-query 的基础上通过在末尾附加 .lean() 来禁用此行为。如果您的查询是 returning “大量”文档，这确实可以提高您的速度。您应该从上面的 link 阅读更多关于 lean() 的信息。

查询优化：

When evaluating the clauses in the $or expression, MongoDB either performs a collection scan or, if all the clauses are supported by indexes, MongoDB performs index scans. That is, for MongoDB to use indexes to evaluate an $or expression, all the clauses in the $or expression must be supported by indexes. Otherwise, MongoDB will perform a collection scan.

https://www.mongodb.com/docs/manual/reference/operator/query/or/#-or-clauses-and-indexes

您共享的查询如下所示：

const involvedUsers = await User.find({
  $or: [
    { interested_cities: { $in: citiesArr } },
    { phone_number: { $in: phonesArr } },
  ],
}, { _id: 1, "push_notification_tokens_firebase.fcm_token": 1 });

根据以上信息，您需要创建以下两个索引：

userSchema.index({ interested_cities: 1 });
userSchema.index({ phone_number: 1 });

这样，mongo 将能够“知道”哪些文档是相关的，在磁盘上找到它们，提取您的投影（“_id”和“push_notification_tokens_firebase.fcm_token”）和return它。

优化的更进一步是创建以下索引而不是上面的索引：

userSchema.index({ interested_cities: 1, _id: 1, "push_notification_tokens_firebase.fcm_token": 1 });
userSchema.index({ phone_number: 1, _id: 1, "push_notification_tokens_firebase.fcm_token": 1 });

这样，mongo 将从索引中获得完成查询所需的所有信息，这意味着它永远不会访问磁盘来获取文档。

您可以通过运行 <your-query>.explain('executionStats') 确认 totalDocsExamined 是 0。

在此处阅读有关 executionStats 的更多信息： https://www.mongodb.com/docs/manual/reference/explain-results/#mongodb-data-explain.executionStats

希望对您有所帮助！

MongoDB 数组的性能问题

MongoDB Performance Issue with arrays

mongodb

mongoose

mongodb-query

query-optimization

mongoose-schema