Andy Pan

> is go using the same epoll instance accross all threads? that might be the underlying problem, most high-throughput applications (nginx, envoy, netty) create several instances (usually one per thread...

> using multiple `epoll` instances would mean that connections or fds would now be bound to a single thread? does this means that it could be possible for connection imbalances...

> I wrote a workaround that does not invoke netpoll at all, instead it just makes raw syscalls and throughput improved by 5x Just to make sure I don't misread...

> Is anybody interested in seeing whether [go.dev/cl/564197](https://go.dev/cl/564197) fixes the problem? To be clear, I'm not going to submit it unless I have some reason to think that it helps....

> I might have some cycles this week to test my reproducer. Great! Thanks! @bwerthmann

I ran into this issue today, and it seems to be recent and there is still no answer to it. While waiting for the official team to fix this, there...

怎么还是用的这个模板。。。算了，看起来相关的信息应该够了。

看起来像是你的 freebsd 系统没有 `kqueue`, 但是你的系统版本又是 11.2，按道理应该是有的。你的 freebsd 是安装在物理机还是虚拟机上的？ 运行一下 `man 2 kevent` 命令看看你的系统里有没有这个系统调用。

你是在服务器上进行虚拟化还是桌面电脑的虚拟机软件？这个问题比较奇怪，可能是虚拟机的问题，我以前在 freebsd 云服务器上测试过是没问题的。

There seems to be a lot of CI failures, please fix that.