Phoenix/Cowboy Request Life-cycle especially When to refresh browser!

controller

  def process_test(conn, _) do
    Plug.Conn.configure_session(conn, renew: true)
    Logger.debug("Controler Process : #{inspect self()}")

    make_monitor_process()

    # IO.inspect conn
    Process.sleep(10_000)

    text conn, "#{inspect self()} : Done"
  end

  def make_monitor_process do
    target_pid = self()

    spawn(fn ->
      Logger.debug("Monitor Process : #{inspect self()} , target : #{inspect target_pid}")
      Process.flag(:trap_exit, true)
      Process.link(target_pid)

      receive do
        msg -> Logger.debug("Received : #{inspect msg}")
      end
    end)
  end

open your browser and put “localhost:4000/test” on address box and push enter key,
and then refresh your browser before 10 seconds. (just refresh as soon as push enter key)

you’ll see below.

iex(5)> [info] GET /test
[debug] Processing with TestWeb.PageController.process_test/2
  Parameters: %{}
  Pipelines: [:browser, :not_authenticated]
[debug] Controler Process : #PID<0.1335.0>
[debug] Monitor Process : #PID<0.1345.0> , target : #PID<0.1335.0>

*> [debug] Received : {:EXIT, #PID<0.1335.0>, :shutdown}*

[info] GET /test
[debug] Processing with TestWeb.PageController.process_test/2
  Parameters: %{}
  Pipelines: [:browser, :not_authenticated]
[debug] Controler Process : #PID<0.1347.0>
[debug] Monitor Process : #PID<0.1357.0> , target : #PID<0.1347.0>
[info] Sent 200 in 10004ms

I am guessing that cowboy killed previous request process
whenever it receives next request before sending response for first one.

my question is that where I can find cowbow or something else’s source
that affect killing process I mentioned above?

and is there any ways to config that kind of thing like waiting for processing request instead of killing it?

Thanks.

What is it that you want to continue doing and why do you want to keep doing it once the connection has been closed? Where’s the result of this supposed to end up?

I just want to know how it works and who makes it get killed.

Received : {:EXIT, #PID<0.1335.0>, :shutdown}*

I think it’s sophisticated way that killing unfinished process just after receiving new process.
and wanna know the drills

There’s no previous/next process to cowboy – it just deals with requests one at a time. If a request is closed by the client (when you hit reload, the active request is closed) the process handling that request will be shut down. There’s noone to receive a response to the request anymore anyways. This happens completely independant to any other request cowboy might be handling.

You can probably see this by closing your browser instead of hitting reload. This should close the active request and not start a new one.

Thanks @LostKobrakai.

I get that cowboy kill current request process when the connection between browser and cowboy has been destroyed.