首页 > 解决方案 > 为什么多个sql查询顺序运行而不是同时运行?

问题描述

我同时在单独的 go 例程中运行 mysql“选择”查询。然而,mysql 服务似乎会收集这些查询,按顺序(非同时)运行它们,然后同时返回所有结果集,并且只有在所有查询都运行后才返回。

我的问题是:
1. 为什么这些查询是按顺序运行而不是同时运行的?
2. 为什么mysql要等到所有查询都运行完才同时返回每个查询的结果集(即使每个单独的结果集属于不同的goroutine并且据说也使用单独的连接)?

另一件事:当我设置“SetMaxOpenConns(2)”时,它会同时返回两个结果集。如果设置为 3,则同时返回 3 个结果集。但是,它们仍然始终按顺序运行。

有人知道这里发生了什么吗?

package main

import (
    "database/sql"
    "fmt"
    "sync"
    "time"

    _ "github.com/go-sql-driver/mysql"
)

var database *sql.DB

func init() {
    var err error
    databaseURI := "root:toor@tcp(192.168.200.10:3306)/ahc"
    database, err = sql.Open("mysql", databaseURI)
    if err != nil {
        fmt.Println(err)
    } else {
        fmt.Println("DB Connection Established")
        //database.SetMaxIdleConns(0)
        //database.SetMaxOpenConns(2)
    }
}

func test(device string, wg *sync.WaitGroup) {
    fmt.Println("Thread: " + device + " started")
    start := time.Now()

    var count string

    //using go-sql-driver
    sqlStatement, err := database.Prepare("select count(cpeName) from report where errMessage <> \"ok\" and cpeName = ? and jobID = ?")
    if err != nil {
        fmt.Println(err)
    }
    defer sqlStatement.Close()
    err = sqlStatement.QueryRow(device, "11534").Scan(&count)
    sqlStatement.Close()


    duration := time.Since(start)
    fmt.Println("Thread: " + device + " Duration: " + duration.String() + "\n")
    wg.Done()
}

func main() {
    var wg sync.WaitGroup
    var deviceList = []string{"xx-swrk-ca-gen-s-002", "xx-leus-ca-ust-ap-068", "xx-sgvn-ca-lug-ap-004", "xx-swrk-ca-vez-s-005", "xx-swrk-ca-vez-ap-006",    "xx-leus-ca-ust-ap-065", "xx-leus-ca-ust-ap-073", "xx-leus-ca-ust-ap-076", "xx-leus-ca-ust-ap-077", "xx-swrk-ca-gen-s-001"}
    total := time.Now()
    for _, device := range deviceList {
        wg.Add(1)
        go test(device, &wg)
    }
    wg.Wait()
    duration := time.Since(total)
    fmt.Println("\n\nTotal: Duration: " + duration.String() + "\n")
}

这是输出

DB Connection Established
Thread: xx-leus-ca-ust-ap-068 started
Thread: xx-sgvn-ca-lug-ap-004 started
Thread: xx-leus-ca-ust-ap-065 started
Thread: xx-leus-ca-ust-ap-073 started
Thread: xx-swrk-ca-vez-ap-006 started
Thread: xx-swrk-ca-vez-s-005 started
Thread: xx-leus-ca-ust-ap-076 started
Thread: xx-leus-ca-ust-ap-077 started
Thread: xx-swrk-ca-gen-s-002 started
Thread: xx-swrk-ca-gen-s-001 started
Thread: xx-leus-ca-ust-ap-076 Duration: 7.291656143s

Thread: xx-swrk-ca-gen-s-002 Duration: 7.304134404s

Thread: xx-leus-ca-ust-ap-065 Duration: 7.307958641s

Thread: xx-swrk-ca-vez-s-005 Duration: 7.313591747s

Thread: xx-leus-ca-ust-ap-077 Duration: 7.313992638s

Thread: xx-swrk-ca-vez-ap-006 Duration: 7.314905664s

Thread: xx-swrk-ca-gen-s-001 Duration: 7.320466323s

Thread: xx-leus-ca-ust-ap-073 Duration: 7.322158337s

Thread: xx-leus-ca-ust-ap-068 Duration: 7.324745097s

Thread: xx-sgvn-ca-lug-ap-004 Duration: 7.326001783s



Total: Duration: 7.326096238s

使用 database.SetMaxOpenConns(1) 时,输出如下:

DB Connection Established
Thread: xx-leus-ca-ust-ap-068 started
Thread: xx-swrk-ca-gen-s-001 started
Thread: xx-swrk-ca-vez-ap-006 started
Thread: xx-leus-ca-ust-ap-065 started
Thread: xx-leus-ca-ust-ap-073 started
Thread: xx-swrk-ca-gen-s-002 started
Thread: xx-leus-ca-ust-ap-077 started
Thread: xx-sgvn-ca-lug-ap-004 started
Thread: xx-leus-ca-ust-ap-076 started
Thread: xx-swrk-ca-vez-s-005 started
Thread: xx-leus-ca-ust-ap-068 Duration: 1.131790286s

Thread: xx-leus-ca-ust-ap-077 Duration: 2.128919333s

Thread: xx-swrk-ca-gen-s-001 Duration: 3.073559464s

Thread: xx-leus-ca-ust-ap-073 Duration: 4.002964333s

Thread: xx-swrk-ca-vez-s-005 Duration: 4.932256684s

Thread: xx-sgvn-ca-lug-ap-004 Duration: 5.853361245s

Thread: xx-swrk-ca-gen-s-002 Duration: 6.785042625s

Thread: xx-leus-ca-ust-ap-065 Duration: 7.705957815s

Thread: xx-swrk-ca-vez-ap-006 Duration: 8.633000734s

Thread: xx-leus-ca-ust-ap-076 Duration: 9.550948572s



Total: Duration: 9.551103129s

标签: gomariadbgoroutine

解决方案


查看它们是并行运行还是串行运行的一种简单技术:让每个连接都执行

SELECT SLEEP(1);

如果是并行的,则该集合不会超过 1 秒。如果串行,则 N 秒。

如果它们是按顺序运行的,但在所有完成之前没有输出,那将是一个 GO 问题。

如果你的 7.3s 真的是并行的,而 9.5 是串行的,那么这就说明了为什么并行运行东西是不值得的——它们会相互踩踏,并且不会更快地完成。冲突可能是 CPU 或 I/O 或互斥锁或网络或其他东西;这取决于查询。(我的sleep测试是非常无创的,但需要的时间是可以预测的。)


推荐阅读