出现了较多的TIME-WAIT

[JJFLYCLOUD]

通过rathole穿透了一个socks5服务
在服务端执行 ss -na|grep x.x.x.x(对应客户端的公网IP)，低负载情况下(几十KB)出现了七十多个TIME-WAIT，而ESTAB只有十几个
使用frp则只有一个 ESTAB

因为负载比较低所以实际使用没什么区别，不知道这是rathole的特性，还是我的网络本身有问题

[yujqiao]

This is a fair point. we should avoid unnecessary timewait on the server side. the heartbeat should be bidirectional so that client can close connection actively to avoid timewait on server.

[leptonyu]

copy_bidirectional cannot handle such case.

1. Enable keep alive.
2. Regular check copy_bidirectional in case it's hanging.