Question

在for循环中依次执行插入，更新，查询和删除500,000次，运行十秒后发生“连接太多”，即使我将MySQL的最大连接设置为10,000。运行“show processlist”时，睡眠状态连接的数量很快就会增加到10,000左右。（但是在单独执行CURD时很好，很快完成1000,000次循环）。

只执行for循环一次正常工作：2017-05-16 17:10:26总费用时间为5.241008ms。

下面是代码：

func insertUnis(db *sql.DB, name string){
    stmt_insert, err := db.Prepare("insert into memDB1 values(?,?,'world.cnworld.cnworld.cnworld.cn',?,'ee02:123::af01:9231:df18:8998:ee02:123::af01:9231:df18:8998:abcd','2017070035|2703258|1943|0|0|10.121.205.248|088:abcd:e02:123::/64','Framed-Interface-Id`varchar(64)COLLATEutf8_binNOTNULLdefaultthah','varchar(64)COLLATEutf8_binNOTNULLdefault0xBF019231DF18899934hdas','Call-From-Id`varchar(32)COLLATEd',?,'Current-Bandwidt','NAS-Port-Idvarchar(256)COLLATEutf8_binNOTNULLdefaultsrunk 6/0/36:33.351 0/0/0/0/0/0NAS-Port-Idvarchar(256)COLLATEutf8_binNOTNULLdefaultsrunk 6/0/36:33.351 0/0/0/0/0/0NAS-Port-Idvarchar(256)COLLATEutf8_binNOTNULLdefaultsrunk 6/0/36:33.351 0/0/0/0/0/0sefiuhu',48204,57239,2458787490,2458787490,2458787490,'Static-Bandwidthvhar(24)','Bas-Bandwidth` varchar(64) COLLATE utf8_bin NOT NULL defaultegse','Session-Id varchar(64) COLLATE utf8_bin NOT NULL defaultlasefjie','Coa-Session-Id` varchar(64) COLLATE utf8_bin NOT NULL defaultsdf','macaddr`varchar(32) COLLATE utf8',?,65535,45656,'Reseverd1` varchar(32) COLLATE u',4294967292)")
    if err != nil {
        Log(LL_ERROR, "stmt_insert Prepare err:", err.Error(), ".")
    }
    defer stmt_insert.Close()
    stmt_update, err := db.Prepare("update memDB1 set user_name=? where session_key=?")
    if err != nil {
        Log(LL_ERROR, "stmt_update Prepare err:", err.Error(), ".")
    }   
    defer stmt_update.Close()
    stmt_del, err := db.Prepare("delete from memDB1 where session_key=?")
        if err != nil {
            Log(LL_ERROR, "stmt_del Prepare err:", err.Error(), ".")
    }
    defer stmt_del.Close() 
    for i:=0;i<500000;i++{
        session_key:=string(Krand(64,3))
        user_name := string(Krand(64,3))
        frame_ip := string(Krand(32,3))     
        Nas_IP := string(Krand(32,3))   
        nat_ip := string(Krand(32,3))   

        _, err = stmt_insert.Exec(session_key,user_name,frame_ip,Nas_IP,nat_ip)
        if err != nil {
            Log(LL_ERROR, "stmt_insert Exec err:", err.Error(), ".")
        }

        _, err = stmt_update.Exec(string(Krand(64,3)), session_key)
        if err != nil {
            Log(LL_ERROR, "stmt_update Exec err:", err.Error(), ".")
        }
        sql := "SELECT * FROM memDB1 where session_key='" + session_key+"'"
        row, err := db.Query(sql)
        defer row.Close()
        if err != nil {
            Log(LL_ERROR, "Query err:", err.Error(), ".")
        }
        _, err = stmt_del.Exec(session_key)
         if err != nil {
            Log(LL_ERROR, "stmt_del Exec err:", err.Error(), ".")
        }
    }
    common_package.WgDay.Done()
    return
}

MySQL进程列表：

mysql> show full processlist;
...............
| 1292138 | cid  | localhost:46799 | ciddb | Sleep   |    0 |          | NULL                  |
| 1292139 | cid  | localhost:46800 | ciddb | Sleep   |    0 |          | NULL                  |
| 1292140 | cid  | localhost:46801 | ciddb | Sleep   |    0 |          | NULL                  |
| 1292141 | cid  | localhost:46802 | ciddb | Sleep   |    0 |          | NULL                  |
+---------+------+-----------------+-------+---------+------+----------+-----------------------+
8719 rows in set (0.04 sec)

MySQL错误消息：

2017-05-16 16:37:56 <LL_ERROR> stmt_insert Exec err:Error 1040: Too many connections. 
2017-05-16 16:37:56 <LL_ERROR> stmt_update Exec err:Error 1040: Too many connections. 
2017-05-16 16:37:56 <LL_ERROR> Query err:Error 1040: Too many connections. 2017-05-16 16:37:56 <LL_ERROR> stmt_del Exec err:Error 1040: Too many connections. 
2017-05-16 16:37:56 <LL_ERROR> stmt_insert Exec err:Error 1040: Too many connections. 
2017-05-16 16:37:56 <LL_ERROR> stmt_update Exec err:Error 1040: Too many connections

MySQL设置：max_connections 10050 set global wait_timeout = 30; set global interactive_timeout = 30;

它是一个内存表，有5个索引：KEY index1 (session_key), KEY index2 (Nas-IP), KEY index3 (user_name), KEY index4 (nat_ip), KEY index5 (frame_ip)

Answer 1

你在一个循环中有一个延迟，它只是将row.Close函数推送到堆栈上，并在函数返回时执行。你可能想要这样的东西：

for i:=0;i<500000;i++{
    ...
    sql := "SELECT * FROM memDB1 where session_key='" + session_key+"'"
    row, err := db.Query(sql)
    if err != nil {
        Log(LL_ERROR, "Query err:", err.Error(), ".")
        continue
    }
    row.Close()
    ...
}

支持文档： https://blog.golang.org/defer-panic-and-recover

MySQL连接错误（使用golang）：连接太多，运行show processlist的睡眠连接太多（8000多个）

1 个答案: