Eventmachine talk - Jan 2008 - stop_event_loop doesn''t let connection send data

I got a server running:

EventMachine.run do
   EventMachine.start_server(@host, @port, Connection)
end

When I call:
EventMachine.stop_event_loop

it looks like it''s letting the connections finish their work before
returning and stopping the server, but doesn''t seem like the data is
sent back the the client.

Is that the correct way to stop the server without dropping any
connection?

Thx,
Marc

On Jan 26, 2008 11:38 PM, Marc-Andr? Cournoyer <macournoyer at gmail.com>
wrote:
> I got a server running:
>
> EventMachine.run do
>   EventMachine.start_server(@host, @port, Connection)
> end
>
> When I call:
> EventMachine.stop_event_loop
>
> it looks like it''s letting the connections finish their work
before
> returning and stopping the server, but doesn''t seem like the data
is
> sent back the the client.
>
> Is that the correct way to stop the server without dropping any
> connection?
>

No, just call EM::stop.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080127/014aa736/attachment-0001.html

same behaviour w/ EM::stop, the connections keep processing but the
data is not sent back to the client.

On 27-Jan-08, at 1:49 PM, Francis Cianfrocca wrote:
> On Jan 26, 2008 11:38 PM, Marc-Andr? Cournoyer
> <macournoyer at gmail.com> wrote:
> I got a server running:
>
> EventMachine.run do
>   EventMachine.start_server(@host, @port, Connection)
> end
>
> When I call:
> EventMachine.stop_event_loop
>
> it looks like it''s letting the connections finish their work
before
> returning and stopping the server, but doesn''t seem like the data
is
> sent back the the client.
>
> Is that the correct way to stop the server without dropping any
> connection?
>
>
> No, just call EM::stop.
>
>
> _______________________________________________
> Eventmachine-talk mailing list
> Eventmachine-talk at rubyforge.org
> http://rubyforge.org/mailman/listinfo/eventmachine-talk
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080127/552fa318/attachment.html

On Jan 26, 2008 11:38 PM, Marc-Andr? Cournoyer <macournoyer at gmail.com>
wrote:
> I got a server running:
>
> EventMachine.run do
>   EventMachine.start_server(@host, @port, Connection)
> end
>
> When I call:
> EventMachine.stop_event_loop
>
> it looks like it''s letting the connections finish their work
before
> returning and stopping the server, but doesn''t seem like the data
is
> sent back the the client.
>


Tell me what you mean by "connections finish their work." EM#stop
works by
scheduling a halt to the reactor loop, after the current pass. It''s
entirely
possible for some already-scheduled events to be passed to your Ruby code
after a call to EM#stop.

I suppose it would be possible to add a version of EM#stop that would first
drain all outbound data. Would that help you?

Why are you calling EM#stop in the first place? Does your design rely on
starting, stopping and then restarting the reactor? If so, then there may be
a more effective design.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080128/e3669884/attachment.html

On 28 Jan 2008, at 01:49, Marc-Andr? Cournoyer wrote:
> same behaviour w/ EM::stop, the connections keep processing but the
> data is not sent back to the client.
$stdout.sync = true ?

here''s my code to reproduce:

=====require ''rubygems''
require ''eventmachine''
require ''open-uri''

class Connection < EventMachine::Connection
   def receive_data(data)
     EventMachine.stop
     send_data "HTTP/1.1 200 OK\r\nContent-Type: text/plain\r\nContent-
Length: 3\r\n\r\nok!"
     close_connection_after_writing
   end
end

fork do
   sleep 1
   puts "Received " +
open(''http://0.0.0.0:5000/'').read
end

EventMachine.run do
   puts ''Starting server ...''
   EventMachine.start_server ''0.0.0.0'', 5000, Connection
end
=====
I get:
ruby stop.rb
Starting server ...
/usr/local/lib/ruby/1.8/net/protocol.rb:133:in `sysread'': end of file
reached (EOFError)
	from /usr/local/lib/ruby/1.8/net/protocol.rb:133:in `rbuf_fill''
	from /usr/local/lib/ruby/1.8/timeout.rb:56:in `timeout''
	from /usr/local/lib/ruby/1.8/timeout.rb:76:in `timeout''
	from /usr/local/lib/ruby/1.8/net/protocol.rb:132:in `rbuf_fill''
	from /usr/local/lib/ruby/1.8/net/protocol.rb:116:in `readuntil''
	from /usr/local/lib/ruby/1.8/net/protocol.rb:126:in `readline''
	from /usr/local/lib/ruby/1.8/net/http.rb:2029:in `read_status_line''
	from /usr/local/lib/ruby/1.8/net/http.rb:2018:in `read_new''
	 ... 10 levels...
	from /usr/local/lib/ruby/1.8/open-uri.rb:30:in `open''
	from stop.rb:15
	from stop.rb:13:in `fork''
	from stop.rb:13


I would like to let the connection send the data before closing.

Is it possible?

On 28-Jan-08, at 8:37 AM, Francis Cianfrocca wrote:
> On Jan 26, 2008 11:38 PM, Marc-Andr? Cournoyer
> <macournoyer at gmail.com> wrote:
> I got a server running:
>
> EventMachine.run do
>   EventMachine.start_server(@host, @port, Connection)
> end
>
> When I call:
> EventMachine.stop_event_loop
>
> it looks like it''s letting the connections finish their work
before
> returning and stopping the server, but doesn''t seem like the data
is
> sent back the the client.
>
>
>
> Tell me what you mean by "connections finish their work." EM#stop
> works by scheduling a halt to the reactor loop, after the current
> pass. It''s entirely possible for some already-scheduled events to
be
> passed to your Ruby code after a call to EM#stop.
>
> I suppose it would be possible to add a version of EM#stop that
> would first drain all outbound data. Would that help you?
>
> Why are you calling EM#stop in the first place? Does your design
> rely on starting, stopping and then restarting the reactor? If so,
> then there may be a more effective design.
> _______________________________________________
> Eventmachine-talk mailing list
> Eventmachine-talk at rubyforge.org
> http://rubyforge.org/mailman/listinfo/eventmachine-talk
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080128/ca1a66c5/attachment-0001.html

On Jan 28, 2008 5:37 PM, Marc-Andr? Cournoyer <macournoyer at gmail.com>
wrote:
> here''s my code to reproduce:
> =====> require ''rubygems''
> require ''eventmachine''
> require ''open-uri''
>
> class Connection < EventMachine::Connection
>   def receive_data(data)
>     EventMachine.stop
>     send_data "HTTP/1.1 200 OK\r\nContent-Type:
> text/plain\r\nContent-Length: 3\r\n\r\nok!"
>     close_connection_after_writing
>   end
> end
>
> fork do
>   sleep 1
>   puts "Received " +
open(''http://0.0.0.0:5000/'').read<http://0.0.0.0:5000/%27%29.read>
> end
>
> EventMachine.run do
>   puts ''Starting server ...''
>   EventMachine.start_server ''0.0.0.0'', 5000, Connection
> end
> =====>
>
If you want to stop the reactor after serving only a single connection, then
put the EM#stop call in the unbind method of the connection object.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080128/8af26dfd/attachment.html

On 28-Jan-08, at 6:04 PM, Francis Cianfrocca wrote:
>
> If you want to stop the reactor after serving only a single
> connection, then put the EM#stop call in the unbind method of the
> connection object.
no, this was just to show you what I''m trying to accomplish.
It''s for
my web server Thin, I don''t wanna drop any request that is currently
processing.
EventMachine.stop is called when a INT signal is caught.
Remove the stop from my example and imagine it''s stopped from another
thread.

On Jan 28, 2008 7:06 PM, Marc-Andr? Cournoyer <macournoyer at gmail.com>
wrote:
> On 28-Jan-08, at 6:04 PM, Francis Cianfrocca wrote:
> >
> > If you want to stop the reactor after serving only a single
> > connection, then put the EM#stop call in the unbind method of the
> > connection object.
>
> no, this was just to show you what I''m trying to accomplish.
It''s for
> my web server Thin, I don''t wanna drop any request that is
currently
> processing.
> EventMachine.stop is called when a INT signal is caught.
> Remove the stop from my example and imagine it''s stopped from
another
> thread.
>
>
>
So when you catch INT, you want to complete all the pending transactions
before you terminate the program?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080128/c92e2ef2/attachment.html

On 28-Jan-08, at 8:34 PM, Francis Cianfrocca wrote:
>
> So when you catch INT, you want to complete all the pending
> transactions before you terminate the program?
>
Exactly!

Is there something builtin EM to do this or do I have to implement
this logic myself?

On Jan 28, 2008 8:50 PM, Marc-Andr? Cournoyer <macournoyer at gmail.com>
wrote:
>
> On 28-Jan-08, at 8:34 PM, Francis Cianfrocca wrote:
> >
> > So when you catch INT, you want to complete all the pending
> > transactions before you terminate the program?
> >
>
> Exactly!
>
> Is there something builtin EM to do this or do I have to implement
> this logic myself?
>
>
You have several choices.

The key question is: when you receive SIGINT, do you want to force
connections to close immediately, or do you want them to finish at their own
pace?

In either case, you''ll need either a class variable or a global array
that
remembers all accepted connections. Then, when you want to run them down,
all you have to do is keep track of when they end by adding some code to the
#unbind method.

Your sigint handler should NOT do all this work. Rather, do something like
this:

trap("INT") {
  EM.next_tick { your_function_to_stop_the_reactor }
}

You also need to call EM#stop_tcp_server to keep new connections from being
accepted while the existing ones drain out.

Feel free to contact me off list and I can help you get this written.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080129/e828f309/attachment.html

thx Francis that worked perfectly!

On 29-Jan-08, at 8:25 AM, Francis Cianfrocca wrote:
> On Jan 28, 2008 8:50 PM, Marc-Andr? Cournoyer
> <macournoyer at gmail.com> wrote:
>
> On 28-Jan-08, at 8:34 PM, Francis Cianfrocca wrote:
> >
> > So when you catch INT, you want to complete all the pending
> > transactions before you terminate the program?
> >
>
> Exactly!
>
> Is there something builtin EM to do this or do I have to implement
> this logic myself?
>
>
> You have several choices.
>
> The key question is: when you receive SIGINT, do you want to force
> connections to close immediately, or do you want them to finish at
> their own pace?
>
> In either case, you''ll need either a class variable or a global
> array that remembers all accepted connections. Then, when you want
> to run them down, all you have to do is keep track of when they end
> by adding some code to the #unbind method.
>
> Your sigint handler should NOT do all this work. Rather, do
> something like this:
>
> trap("INT") {
>   EM.next_tick { your_function_to_stop_the_reactor }
> }
>
> You also need to call EM#stop_tcp_server to keep new connections
> from being accepted while the existing ones drain out.
>
> Feel free to contact me off list and I can help you get this written.
> _______________________________________________
> Eventmachine-talk mailing list
> Eventmachine-talk at rubyforge.org
> http://rubyforge.org/mailman/listinfo/eventmachine-talk
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/eventmachine-talk/attachments/20080130/82d72185/attachment.html