在Golang和MongoDB中使用$ lookup和$ unwind的请求很慢

时间:2018-06-19 14:30:31

标签: mongodb go mgo

我的模型:活动可以由具有一个或多个 Boost Campaign 增强。

我正在尝试通过以下请求获取所有被提升 Event

// Boosted ...
func (dao *campaignDAO) Boosted() ([]*models.Event, error) {
    // Clone the session
    session := dao.session.Clone()
    defer session.Close()

    // Get the time
    now := time.Now()

    // Create the pipe
    pipe := session.DB(shared.DatabaseNamespace).C("events").Pipe([]bson.M{
        {
            "$lookup": bson.M{
                "from":         "event_boosts",
                "localField":   "_id",
                "foreignField": "_event_id",
                "as":           "boost",
            },
        },
        {"$unwind": "$boost"},
        {
            "$match": bson.M{
                "boost.is_published": true,               // Boost is active
                "boost.start_date":   bson.M{"$lt": now}, // now is between start and end
                "boost.end_date":     bson.M{"$gt": now}, // now is between start and end
            },
        },
        {
            "$lookup": bson.M{
                "from":         "campaigns",
                "localField":   "boost._campaign_id",
                "foreignField": "_id",
                "as":           "campaign",
            },
        },
        {"$unwind": "$campaign"},
        {
            "$match": bson.M{
                "campaign.is_published": true, // Attached campaign is active
            },
        },
    })

    var result []*models.Event
    err := pipe.All(&result)
    if err != nil {
        return nil, err
    }

    return result, nil
}

但是此请求需要3秒钟。 这是我在广告系列中拥有的索引:

// NewCampaignDAO returns a new CampaignDAO
func NewCampaignDAO(session *mgo.Session) dao.CampaignDAO {
    // Set the collection
    col := session.DB(shared.DatabaseNamespace).C("campaigns")

    // Set the indexes
    col.EnsureIndexKey("start_date")
    col.EnsureIndexKey("end_date")
    col.EnsureIndexKey("created_by")
    col.EnsureIndexKey("is_published")

    return &campaignDAO{
        session:    session,
        collection: "campaigns",
    }
}

事件索引:

// NewEventDAO returns a new EventDAO
func NewEventDAO(session *mgo.Session) dao.EventDAO {
    // Set the collection
    col := session.DB(shared.DatabaseNamespace).C("events")

    // Set the indexes
    col.EnsureIndexKey("old_id")
    col.EnsureIndexKey("_parent_id")
    col.EnsureIndexKey("_location_id")
    col.EnsureIndexKey("price")
    col.EnsureIndexKey("name")
    col.EnsureIndexKey("category")
    col.EnsureIndexKey("start_date")
    col.EnsureIndexKey("end_date")
    col.EnsureIndexKey("is_recurrent")
    col.EnsureIndexKey("is_published")
    col.EnsureIndexKey("is_proposed")
    col.EnsureIndexKey("tags")
    col.EnsureIndexKey("price", "date", "name")

    return &eventDAO{
        session:    session,
        collection: "events",
    }
}

还有MongoDB的日志:

2018-06-19T13:22:53.465+0000 I COMMAND  [conn506] command clutch.event_boosts command: aggregate { aggregate: "events", pipeline: [ { $lookup: { as: "boost", from: "event_boosts", localField: "_id", foreignField: "_event_id" } }, { $unwind: "$boost" }, { $match: { boost.is_published: true, boost.start_date: { $lt: new Date(1529414570196) }, boost.end_date: { $gt: new Date(1529414570196) } } }, { $lookup: { from: "campaigns", localField: "boost._campaign_id", foreignField: "_id", as: "campaign" } }, { $unwind: "$campaign" }, { $match: { campaign.is_published: true } } ], cursor: {} } planSummary: COLLSCAN keysExamined:0 docsExamined:12936 cursorExhausted:1 numYields:121 nreturned:1 reslen:1149 locks:{ Global: { acquireCount: { r: 52018 } }, Database: { acquireCount: { r: 26009 } }, Collection: { acquireCount: { r: 26008 } } } protocol:op_query 3268ms

我没办法去改进。

编辑:另外,我想知道是否可以通过在 event_boosts 集合上启动请求,然后在“事件”中查找来改善请求。

EDIT2:添加Mongo版本。

MongoDB shell version v3.4.6
git version: c55eb86ef46ee7aede3b1e2a5d184a7df4bfb5b5
OpenSSL version: OpenSSL 1.0.1t  3 May 2016
allocator: tcmalloc
modules: none
build environment:
    distmod: debian81
    distarch: x86_64
    target_arch: x86_64

1 个答案:

答案 0 :(得分:0)

这大概是我想提供的帮助。由于我没有示例数据,因此未经测试。另外,由于我不了解Go,它的Go语法可能有些古怪。 ;)但是,我相对确定$match $lookup内的pipeline语句将利用可用索引,而在您的查询中,您在{ {1}}个有效地使索引无用的东西。

$unwind