Goroutine leak when dial err in go1.8

[qw4990]

Issue description

Goroutine leak when dial err in go1.8

Example code

package main

import (
	"fmt"
	"log"
	"net/http"
	_ "net/http/pprof"
	"time"

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

func main() {
	go func() {
		log.Println(http.ListenAndServe("localhost:6060", nil))
	}()

	for i := 0; i < 100; i++ {
		var driver mysql.MySQLDriver
		_, err := driver.Open("USER:PASSWORD_HERE@wrong(127.0.0.1:3306)/DATABASE?charset=utf8&parseTime=True&loc=Local")
		fmt.Println(err)
	}

	time.Sleep(time.Hour * 5)
}

Error log

When you run the code above and access http://127.0.0.1:6060/debug/pprof/goroutine?debug=1;

You can see there are 100 leaked goroutines:

100 @ 0x102df2a 0x103d014 0x103bc7c 0x132e738 0x1059b61
#	0x132e737	github.com/go-sql-driver/mysql.(*mysqlConn).startWatcher.func1+0x287	/Users/zhangyuanjia/Work/src/github.com/go-sql-driver/mysql/connection_go18.go:189

Configuration

Driver version (or git SHA): 8fefef0

Go version: go1.8.1 darwin/amd64

Server version: MySQL 5.6

Server OS: OS X

[qw4990]

Fixed in #627

[neyao]

Hi @qw4990
I am a golang engeer in Alibaba. We found our procss hang on execute sql, looks like the same issue as this one.

enviroment:
go version go1.9.2 linux/amd64
CentOS 6.6 LInux 4.4.88-1.el7.elrepo.x86_64

We found our process hang in sql execute, and there are lots of below logs in pprof full stack trace:

goroutine 340 [select, 167 minutes]:
github.com/go-sql-driver/mysql.(*mysqlConn).startWatcher.func1(0xc420173560, 0xc4208640c0, 0xc4203bcde0)
	github.com/go-sql-driver/mysql/connection_go18.go:182 +0xf8
created by github.com/go-sql-driver/mysql.(*mysqlConn).startWatcher
	github.com/go-sql-driver/mysql/connection_go18.go:179 +0xce

There is 452 times such logs in full stack dump logs, and our MaxOpenConns is 512

[methane]

We found our process hang in sql execute,

What is stacktrace for them?

and there are lots of below logs in pprof full stack trace:

It is normal, not a hung.

[neyao]

Thank you methane, our problem solved!

We finally found the reason why the process hung, it was a rookie mistake, we used mysql.Query() to execute update sql and rows not closed, which is exacly the "BAD" usage of this article: http://go-database-sql.org/modifying.html

There was lots of this kind of stacktrace like that:
goroutine 337 [select, 4 minutes]: database/sql.(*Rows).awaitDone(0xc420352980, 0xc8eba0, 0xc42046d340, 0x0, 0x0) golang/go/src/database/sql/sql.go:2606 +0xf3 created by database/sql.(*Rows).initContextClose golang/go/src/database/sql/sql.go:2594 +0x9c

Thanks all anyway.