This message was deleted.

I would like to know that too. Even in dev mode it would be helpful to have an immediate reaction to CTRL_C. However, killing the entire process helps in any case 😃

@dry-elephant-14928 Are we missing some config flag? Or is this kind of a missing feature/bug ?

this is a feature, not a bug 🙂 you can

kill -9

the process if intending to force kill without waiting for participants.

Of course 🙂 But I was expecting some kind of flag to change the value of "force" on this function

func (s *LivekitServer) Stop(force bool) {

Also, its odd why a server needs to wait for the clients forever - I don't remember a behavior like this in other services - but's that ok. Thanks for the answer

the reason why we do not terminate.. is because unlike stateless HTTP servers, real-time sessions are quite disruptive to kill while participants are still exchanging media. When the server dies, all of the clients will experience reconnection at the same time. In Kubernetes environments, you can set a graceful termination timeout, and that is the amount of time LiveKit process will be allowed to "drain" - does not take on new rooms, but will allow current sessions to finish. At the end of that period, Kubernetes will effectively kill -9 the pod