Lesson 09 · Stopping a running service - Build a Process Supervisor

Stopping a running service

To stop a service the supervisor does not yank it away - it politely asks the process to exit by sending SIGTERM, then waits. Today you add Stop, which signals the service's handle and moves it into the Stopping state.

The goal

Stop a running service by sending SIGTERM to its handle and entering Stopping.

Start here - the target

TO DO

Scenario: Stop signals the process and enters Stopping

Givena Running service "web" holding handle h from the runtime

WhenStop("web") is called

Thenthe runtime recorded a SIGTERM delivered to handle h, and the service State is Stopping

Andthe process has not been reaped yet - the service is Stopping, not Stopped (it is waiting for the process to actually exit)

Background

Stopping a process cleanly is a request, not a demand. The supervisor sends SIGTERM - the signal that means “please shut down” - and the process gets to flush its buffers, close its connections, and exit on its own terms. Crucially, the service does not become Stopped the instant you send the signal: the process is still alive, still cleaning up. It is in Stopping, an in-between state that means “we have asked; we are waiting.”

That waiting matters. A well-behaved process exits promptly after SIGTERM, but a stuck one might ignore it entirely - which is exactly why a real supervisor escalates to SIGKILL after a grace period (a lesson you reach at the end). For now, Stop sends the polite signal and records that we are waiting. The moment the process actually exits is a separate event - the reap - and that is next.

Make it work

func (s *Supervisor) Stop(name string) {
  svc, _ := s.Get(name)
  if svc.State != Running { return }
  s.rt.Signal(svc.Handle, SIGTERM) // ask it to exit
  svc.State = Stopping              // waiting for the exit
}

CheckpointDONE

Stopping a service sends SIGTERM and enters Stopping. Commit and stop here.