socket_manager: add feature to share sockets with another server

[daipom]

Another process can take over UDP/TCP sockets without downtime.

server = ServerEngine::SocketManager::Server.share_sockets_with_another_server(path)

This starts a new server that shares all UDP/TCP sockets with the existing server. The old process should stop without removing the file for the socket after the new process starts.

This allows us to replace both the server and the workers with new processes without socket downtime. (The existing live restart feature does not support network servers. We can restart workers without socket downtime, but there is no such way for the network server.)

ref: https://github.com/fluent/fluentd/issues/4622

Limitation

• This feature would not work well if the process opens new TCP ports frequently.
  • If a client listens to a new socket during the sharing process, the new socket may not take over.
  • If this happens ...
  • Safe case: After the old server stops, the new server re-opens it.
  • Unsafe case: The new server tries to re-open it before the old server stops and causes address already in use error.
• The old process must not remove the socket file when it stops.

TODO

• [x] Add tests
• [x] Consider limitations
• [x] Consider exclusive lock

[daipom]

It would be necessary to consider exclusive locks.

[daipom]

https://github.com/treasure-data/serverengine/compare/772d7dfa672038fb0745eee7eb608799ef81b9a1..13b3d398a68d5924e9d5365a302b16c2804a1124

Somehow, the fork process couldn't receive SIGTERM, so I gave up using fork for the tests.

[kenhys]

It may be better to note explicitly the scope of this PR (out of scope live-restart for supervisor, we focus on server <=> worker) https://github.com/treasure-data/serverengine?tab=readme-ov-file#live-restart

[daipom]

Thanks for your review! This PR allows us to restart network servers without socket downtime. I fixed the description of this PR and the commit message.

[daipom]

I'll rebase this.

[daipom]

Thanks for your review!