#14 introduces infinite callback/timeout hell

[serializedowen]

#14 reconnecting happens on try catch block and without proper timer safeguard/cleanup.

observed tcp / eventloop pool overflowing and eventually crashing the whole container.

[soimy]

@serializedowen Currently, the best approach is to implement your own connection watchdog and monitor socket health / heartbeat.

[serializedowen]

@serializedowen Currently, the best approach is to implement your own connection watchdog and monitor socket health / heartbeat.

Given that this is a issue ticket, thread should be scoped in relevant context.

I don't see how "own conn watchdog or w/e" is related?

[soimy]

Given that this is a issue ticket, thread should be scoped in relevant context.

I don't see how "own conn watchdog or w/e" is related?

Not a maintainer, just someone who's facing the same issue as you and fixed this.
Since the SDK is currently broken, any functional workaround is "related".

[serializedowen]

Given that this is a issue ticket, thread should be scoped in relevant context.
I don't see how "own conn watchdog or w/e" is related?

Not a maintainer, just someone who's facing the same issue as you and fixed this. Since the SDK is currently broken, any functional workaround is "related".

thanks for the update, i'll look into a fix for this.

[yilin-zhouyang]

#14重新连接发生在 try catch 块中，并且没有适当的计时器保护/清理。

观察到 TCP/事件循环池溢出，最终导致整个容器崩溃。

遇到了同样的问题，目前我是这样解决的：
1.关闭 SDK 自动重连，业务层接管重连逻辑，增加 watchdog
2.健康检查，SIGTERM 信号处理增加连接清理和进程退出
3.使用 patch-package 对 SDK 源码打补丁（terminate() 替代 close()、重连前销毁旧socket、connect() 入口增加状态检查），确保 npm install 后补丁自动生效