reload error about categraf HOT 18 CLOSED

大概原因是，在重载的时候，先已经stopping关闭了通道，但是指标正好采集回来，正往里面写数据，导致的

from categraf.

复现可以把quitChan分支时间调长

	for {
		select {
		case <-r.quitChan:
			time.Sleep(interval * 10)
			close(r.quitChan)
			return
		default:
			time.Sleep(interval)
			r.gatherOnce()
		}
	}

修复方案，close(r.queue) 从Stop()中移到close(r.quitChan)之前

from categraf.

panic: close of closed channel

goroutine 1 [running]:
flashcat.cloud/categraf/agent.(*InputReader).Stop(...)
/root/categraf/agent/metrics_reader.go:46
flashcat.cloud/categraf/agent.(*Agent).stopMetricsAgent(...)
/root/categraf/agent/metrics_agent.go:90
flashcat.cloud/categraf/agent.(*Agent).Stop(0xc0003eae38)
/root/categraf/agent/agent.go:67 +0xf3
flashcat.cloud/categraf/agent.(*Agent).Reload(0xc000383d58?)
/root/categraf/agent/agent.go:74 +0x57
main.handleSignal(0xc0008f1e60?)
/root/categraf/main.go:94 +0x206
main.runAgent(0x0?)
/root/categraf/main_posix.go:12 +0x28
main.main()
/root/categraf/main.go:69 +0x205

from categraf.

连续两次快速的reload可能引发 panic: close of closed channel
/root/categraf/agent/metrics_reader.go:46
close(r.queue)

from categraf.

@UlricQin 如你所说，之前我被两次测试误解了，之前的判断是错误的。 我之前关闭的issue，在0.1.6测试的，也是quitchan写入时panic

from categraf.

from categraf.

panic是写入r.queue，0.1.13已经修复了。 可以再测试下https://github.com/flashcatcloud/categraf/blob/main/agent/metrics_reader.go#L63

from categraf.

如果没有其他问题，这个issue先close了

from categraf.

并未解决

2022/07/14 22:57:00 main.go:90: I! received signal: hangup
2022/07/14 22:57:00 agent.go:80: I! agent reloading
2022/07/14 22:57:00 agent.go:71: I! agent stopping
panic: send on closed channel

goroutine 1 [running]:
flashcat.cloud/categraf/agent.(*InputReader).Stop(...)
/home/runner/work/categraf/categraf/agent/metrics_reader.go:44
flashcat.cloud/categraf/agent.(*Agent).stopMetricsAgent(...)
/home/runner/work/categraf/categraf/agent/metrics_agent.go:94
flashcat.cloud/categraf/agent.(*Agent).Stop(0xc000bdce38)
/home/runner/work/categraf/categraf/agent/agent.go:73 +0xe5
flashcat.cloud/categraf/agent.(*Agent).Reload(0xc000bdcd58?)
/home/runner/work/categraf/categraf/agent/agent.go:81 +0x57
main.handleSignal(0xc000c2d1c0?)
/home/runner/work/categraf/categraf/main.go:95 +0x206
main.runAgent(0x0?)
/home/runner/work/categraf/categraf/main_posix.go:12 +0x28
main.main()
/home/runner/work/categraf/categraf/main.go:70 +0x209

from categraf.

如之前所说，panic是
/home/runner/work/categraf/categraf/agent/metrics_reader.go:44
写入 quitChan ，所以诡异

from categraf.

./categraf -version
v0.2.2-a7b408226694a3a3bad08f1cfe06fc96cf92cd18

from categraf.

复现方法还是两次reload？ 加载了哪些插件？

from categraf.

没有两次reload，就是一次普通的reload，ping net_response http_response

from categraf.

一次reload 就只写一次quitchan，两次的话，倒是有可能

from categraf.

这个问题，还有出现吗？ @tanxiao1990

from categraf.

老版本一直都有，新版本没更新，如果新版本没有对这段逻辑做处理的话，肯定是存在的。

from categraf.

新版本没有对这块再增加处理逻辑。 我这边复现不出来，好像也没有其他用户再反馈类似的问题。看否有其他线索。

from categraf.

感谢 @tanxiao1990 复现和PR #477
#478

from categraf.