1 |
=head1 Message Passing for the Non-Blocked Mind |
2 |
|
3 |
=head2 Introduction and Terminology |
4 |
|
5 |
This is a tutorial about how to get the swing of the new L<AnyEvent::MP> |
6 |
module. Which allows us to transparently pass messages to our own process and |
7 |
to other process on other or the same host. |
8 |
|
9 |
What kind of messages? Well, basically a message here means a list of |
10 |
Perl strings, numbers, hashes and arrays, mostly everything that can be |
11 |
expressed in a L<JSON> text (as JSON is used by default in the protocol). |
12 |
|
13 |
And next you might ask: between which entities are those messages being "passed"? |
14 |
Basically between C<nodes>, which are basically your applications (as in |
15 |
processes) that use L<AnyEvent::MP> that run either on the same or different |
16 |
hosts. |
17 |
|
18 |
In this tutorial I'll show you how to write a simple chat server based on |
19 |
L<AnyEvent::MP>. |
20 |
|
21 |
=head2 System Requirements |
22 |
|
23 |
Before we can start we have to make sure some things work on your |
24 |
system. First of all the host C<localhost> should resolve to a local |
25 |
IP address. Next you should be able to do TCP over that address. |
26 |
|
27 |
You should of course also make sure that L<AnyEvent> and L<AnyEvent::MP> |
28 |
are installed. But how to do that is out of scope of this tutorial. |
29 |
|
30 |
Then we have to setup a I<shared secret>. For two L<AnyEvent::MP> nodes |
31 |
to be able to communicate with each other and authenticate each other |
32 |
it is necessary to setup a I<shared secret>. For testing you can write a |
33 |
random string followed by a newline into the file C<.aemp-secret> in your |
34 |
home directory: |
35 |
|
36 |
echo "secret123#4blabla_please_pick_your_own" > ~/.aemp-secret |
37 |
|
38 |
Only if the nodes that want to connect to each other have the same I<shared |
39 |
secret> connections will be successful. |
40 |
|
41 |
B<If something does not work as expected, and for example tcpdump shows |
42 |
that the connections are broken up early, you should make sure that ~/.aemp-secret |
43 |
is the same on both hosts/user accounts you are connecting!> |
44 |
|
45 |
=head2 The Chat Client |
46 |
|
47 |
OK, lets start by implementing the "frontend" of the client. We will delay the |
48 |
explanation and the code of the server until we finished the client, as the |
49 |
most complex things actually happen in the client. |
50 |
|
51 |
We will use L<AnyEvent::Handle> to do non-blocking IO read on standard input: |
52 |
|
53 |
#!perl |
54 |
use AnyEvent; |
55 |
use AnyEvent::Handle; |
56 |
|
57 |
sub send_message { |
58 |
die "This is where we will send the messages to the server" |
59 |
. "in the next step of this tutorial.\n" |
60 |
} |
61 |
|
62 |
# make an AnyEvent condition variable for the 'quit' condition |
63 |
# (when we want to exit the client). |
64 |
my $quit_cv = AnyEvent->condvar; |
65 |
|
66 |
my $stdin_hdl = AnyEvent::Handle->new ( |
67 |
fh => \*STDIN, |
68 |
on_read => sub { |
69 |
my ($hdl) = @_; |
70 |
|
71 |
$hdl->push_read (line => sub { |
72 |
my ($hdl, $line) = @_; |
73 |
|
74 |
if ($line =~ /^\/quit/) { # /quit will end the client |
75 |
$quit_cv->send; |
76 |
|
77 |
} else { |
78 |
send_message ($line); |
79 |
} |
80 |
}); |
81 |
} |
82 |
); |
83 |
|
84 |
$quit_cv->recv; |
85 |
|
86 |
This is now a very basic client. Explaining explicitly what |
87 |
L<AnyEvent::Handle> does or what a I<condvar> is all about is out of scope |
88 |
of this document, please consult L<AnyEvent::Intro> or the manual pages |
89 |
for L<AnyEvent> and L<AnyEvent::Handle>. |
90 |
|
91 |
=head2 First Step Into Messaging |
92 |
|
93 |
Now we take a look at L<AnyEvent::MP>. We need to know what to do in |
94 |
C<send_message>. This is an example of how it might look like: |
95 |
|
96 |
... # the use lines from the above snippet |
97 |
|
98 |
use AnyEvent::MP; |
99 |
|
100 |
sub send_message { |
101 |
my ($msg) = @_; |
102 |
|
103 |
snd $server_port, message => $msg; |
104 |
} |
105 |
|
106 |
... # the rest of the above script |
107 |
|
108 |
The C<snd> function is exported by L<AnyEvent::MP>, it stands for 'send a |
109 |
message'. The first argument is the I<port> (a I<port> is something that can |
110 |
receive messages) of the server which will receive the message . How we get it |
111 |
will be explained in the next step. The next arguments of C<snd> are |
112 |
C<message> and C<$msg> are the first two elements of the I<message> (a |
113 |
I<message> in L<AnyEvent::MP> is a be a simple list of values, which can be |
114 |
sent to I<ports>). |
115 |
|
116 |
Thats all fine so far, but how do we get the C<$server_port>? Well, we will |
117 |
need to get the unique I<port id> of the server's port where he wants to |
118 |
receive all the incoming chat messages. A I<port id> is unfortunately a very |
119 |
unique string, which we are unable to know in advance. But L<AnyEvent::MP> |
120 |
supports the concept of 'well known ports', which is basically a port on the |
121 |
server side registered under a well known name. For example, the server has a |
122 |
port for receiving chat messages with a unique I<port id> and registered it |
123 |
under the name C<chatter>. |
124 |
|
125 |
BTW, these "well known port names" should follow similar rules as Perl |
126 |
identifiers, so you should prefix them with your package/module name to |
127 |
make them unique, unless you use them in the main program. |
128 |
|
129 |
As I<messages> can only be sent to a I<port id> and not just to a name we have |
130 |
to ask the server I<node> what I<port id> has the well known port with the |
131 |
name C<chatter>. |
132 |
|
133 |
Another new term, what is a I<node>: The messaging network that can be created with |
134 |
L<AnyEvent::MP> consists of I<nodes>. A I<node> handles all the connection and |
135 |
low level message sending logic for its application. The application in this |
136 |
case is the server. Also every client has/is a I<node>. |
137 |
|
138 |
=head2 Getting The Chatter Port |
139 |
|
140 |
Ok, lots of talk, now some code. Now we will actually get the C<$server_port> |
141 |
from the backend: |
142 |
|
143 |
... |
144 |
|
145 |
use AnyEvent::MP; |
146 |
|
147 |
my $resolved_cv = AnyEvent->condvar; |
148 |
|
149 |
my $client_port = create_port; |
150 |
|
151 |
my $server_node = "localhost:1299#"; |
152 |
|
153 |
snd $server_node, wkp => "chatter", "$client_port", "resolved"; |
154 |
|
155 |
my $server_port; |
156 |
|
157 |
# setup a receiver callback for the 'resolved' message: |
158 |
$client_port->rcv (resolved => sub { |
159 |
my ($client_port, $type, $chatter_port_id) = @_; |
160 |
|
161 |
print "Resolved the server port 'chatter' to $chatter_port_id\n"; |
162 |
$server_port = $chatter_port_id; |
163 |
|
164 |
$resolved_cv->send; |
165 |
1 |
166 |
}); |
167 |
|
168 |
# lets block the client until we resolved the server port. |
169 |
$resolved_cv->recv; |
170 |
|
171 |
# now setup another receiver callback for the chat messages: |
172 |
$client_port->rcv (message => sub { |
173 |
my ($client_port, $type, $msg) = @_; |
174 |
|
175 |
print "chat> $msg\n"; |
176 |
0 |
177 |
}); |
178 |
|
179 |
# send the server a 'join' message: |
180 |
snd $server_port, join => "$client_port"; |
181 |
|
182 |
sub send_message { ... |
183 |
|
184 |
Now that was a lot of new stuff. In order to ask the server and receive an |
185 |
answer we need to have a I<port> where we can receive the answer. |
186 |
This is what C<create_port> will do for us, it just creates a new local |
187 |
port and returns us an object (that will btw. stringify to the I<port id>), |
188 |
that we can use to receive messages. |
189 |
|
190 |
Next thing is the C<$server_node>. In order to refer to another node we need |
191 |
some kind of string to reference it. The I<noderef> is basically a comma |
192 |
separated list of C<host:port> pairs. We assume in this tutorial that the |
193 |
server runs on your localhost at port 1299, this gives us the noderef |
194 |
C<localhost:1299>. |
195 |
|
196 |
Now you might ask what the C<#> at the end in C<$server_node> the above |
197 |
example is about. Well, what I didn't tell you yet is that each I<node> has a |
198 |
default I<port> to receive messages. The default port is the empty string |
199 |
C<"">. The I<default port> of a I<node> also provides some special services for |
200 |
us, for example resolving a well known port to a I<port id>. |
201 |
|
202 |
Now to this line: |
203 |
|
204 |
snd $server_node, wkp => "chatter", "$client_port", "resolved"; |
205 |
|
206 |
We send a message with first element being C<wkp> (standing for 'well known |
207 |
port'). Then the well known port name that we want to resolve to a I<port id>: |
208 |
C<chatter>. And in order for the server node to be able to send us back the |
209 |
resolved I<port id> we have to tell it where to send the result message: The |
210 |
result message will have as it's first argument the string C<resolved> and |
211 |
will be sent to C<$client_port> (the I<port id> of our own just created |
212 |
port). |
213 |
|
214 |
Next comes the receiver for this C<wkp> request. |
215 |
|
216 |
$client_port->rcv (resolved => sub { |
217 |
my ($client_port, $type, $chatter_port_id) = @_; |
218 |
... |
219 |
1 |
220 |
}); |
221 |
|
222 |
This sets up a receiver on our own port for the result message with the first |
223 |
argument being the string C<resolved>. Receivers can match the contents of |
224 |
the messages before actually 'sending' it to the given callback. |
225 |
|
226 |
B<Please note> that the given callback has to return either a true or a false |
227 |
value for indicating whether it is B<done> (true value) or still wants to |
228 |
B<continue> (false value) receiving messages. |
229 |
|
230 |
In this case we tell the C<$client_port> to look into the received messages and |
231 |
look for the string C<resolved> in the first element of the message. If it is |
232 |
found, the given callback will be called with the C<$client_port> as first |
233 |
argument, and the message as the remaining arguments. |
234 |
|
235 |
We name the first element of the message C<$type> in this case. It's a common |
236 |
idiom to code the 'type' of a message into it's first element, this allows for |
237 |
simple matching. |
238 |
|
239 |
The result message will contain the I<port id> of the well known port C<chatter> |
240 |
as next element, and will be put in C<$chatter_port_id>. |
241 |
|
242 |
Next we just assign C<$server_port> and return a 1 (a true value) |
243 |
from the callback. It indicates that we are done and don't want to receive |
244 |
further C<resolved> messages with this callback. |
245 |
|
246 |
Now we continue to the rest of the client by calling C<send> on |
247 |
C<$resolved_cv>. |
248 |
|
249 |
First new step after this is setting up the chat message receiver callback. |
250 |
|
251 |
$client_port->rcv (message => sub { |
252 |
my ($client_port, $type, $msg) = @_; |
253 |
|
254 |
print "chat> $msg\n"; |
255 |
0 |
256 |
}); |
257 |
|
258 |
We assume that all messages that are broadcast to all clients by the server |
259 |
will contain the string C<message> as first element, and the actual message as |
260 |
second element. The callback returns a false value this time, to indicate that |
261 |
it wants to continue receiving messages. |
262 |
|
263 |
Last but not least we actually tell the server to send us |
264 |
the new chat messages from other clients. We do so by sending the |
265 |
message type C<join> followed by our own I<port id>. |
266 |
|
267 |
# send the server a 'join' message: |
268 |
snd $server_port, join => "$client_port"; |
269 |
|
270 |
Then the server knows where to send all the new messages to. |
271 |
|
272 |
=head2 The Completed Client |
273 |
|
274 |
This is the complete client script: |
275 |
|
276 |
#!perl |
277 |
|
278 |
use AnyEvent; |
279 |
use AnyEvent::Handle; |
280 |
use AnyEvent::MP; |
281 |
|
282 |
my $resolved_cv = AnyEvent->condvar; |
283 |
|
284 |
my $client_port = create_port; |
285 |
|
286 |
my $server_node = "localhost:1299#"; |
287 |
|
288 |
snd $server_node, wkp => "chatter", "$client_port", "resolved"; |
289 |
|
290 |
my $server_port; |
291 |
|
292 |
# setup a receiver callback for the 'resolved' message: |
293 |
$client_port->rcv (resolved => sub { |
294 |
my ($client_port, $type, $chatter_port_id) = @_; |
295 |
|
296 |
print "Resolved the server port 'chatter' to $chatter_port_id\n"; |
297 |
$server_port = $chatter_port_id; |
298 |
|
299 |
$resolved_cv->send; |
300 |
1 |
301 |
}); |
302 |
|
303 |
# lets block the client until we resolved the server port. |
304 |
$resolved_cv->recv; |
305 |
|
306 |
# now setup another receiver callback for the chat messages: |
307 |
$client_port->rcv (message => sub { |
308 |
my ($client_port, $type, $msg) = @_; |
309 |
|
310 |
print "chat> $msg\n"; |
311 |
0 |
312 |
}); |
313 |
|
314 |
# send the server a 'join' message: |
315 |
snd $server_port, join => "$client_port"; |
316 |
|
317 |
sub send_message { |
318 |
my ($msg) = @_; |
319 |
|
320 |
snd $server_port, message => $msg; |
321 |
} |
322 |
|
323 |
# make an AnyEvent condition variable for the 'quit' condition |
324 |
# (when we want to exit the client). |
325 |
my $quit_cv = AnyEvent->condvar; |
326 |
|
327 |
my $stdin_hdl = AnyEvent::Handle->new ( |
328 |
fh => \*STDIN, |
329 |
on_read => sub { |
330 |
my ($hdl) = @_; |
331 |
|
332 |
$hdl->push_read (line => sub { |
333 |
my ($hdl, $line) = @_; |
334 |
|
335 |
if ($line =~ /^\/quit/) { # /quit will end the client |
336 |
$quit_cv->send; |
337 |
|
338 |
} else { |
339 |
send_message ($line); |
340 |
} |
341 |
}); |
342 |
} |
343 |
); |
344 |
|
345 |
$quit_cv->recv; |
346 |
|
347 |
=head2 The Server |
348 |
|
349 |
Ok, now finally to the server. What do we need? Well, we need to setup |
350 |
the well known port C<chatter> where all clients send their messages to. |
351 |
|
352 |
Up and into code right now: |
353 |
|
354 |
#!perl |
355 |
|
356 |
use AnyEvent; |
357 |
use AnyEvent::MP; |
358 |
|
359 |
become_public "localhost:1299"; |
360 |
|
361 |
my $chatter_port = create_port; |
362 |
$chatter_port->register ("chatter"); |
363 |
|
364 |
my %client_ports; |
365 |
|
366 |
$chatter_port->rcv (join => sub { |
367 |
my ($chatter_port, $type, $client_port) = @_; |
368 |
|
369 |
print "got new client port: $client_port\n"; |
370 |
|
371 |
$client_ports{$client_port} = 1; |
372 |
0 |
373 |
}); |
374 |
|
375 |
$chatter_port->rcv (message => sub { |
376 |
my ($chatter_port, $type, $msg) = @_; |
377 |
|
378 |
print "message> $msg\n"; |
379 |
|
380 |
snd $_, message => $msg for keys %client_ports; |
381 |
0 |
382 |
}); |
383 |
|
384 |
AnyEvent->condvar->recv; |
385 |
|
386 |
This is all. Looks much easier, doesn't it? I'll explain it only shortly, as |
387 |
we had the discussion of the C<rcv> method in the client part of this tutorial |
388 |
above. |
389 |
|
390 |
First this: |
391 |
|
392 |
become_public "localhost:1299"; |
393 |
|
394 |
This will tell our I<node> to become a I<public> node, which means that it can |
395 |
be contacted via TCP. The first argument should be the I<noderef> the server |
396 |
wants to be reachable at. In this case it's the TCP port 1299 on localhost. |
397 |
|
398 |
Next we basically setup two receivers, one for the C<join> messages and |
399 |
another one for the actual messages of type C<messsage>. |
400 |
|
401 |
In the C<join> message we get the client's port, which we just remember in the |
402 |
C<%client_ports> hash. In the receiver for the message type C<message> we will |
403 |
just iterate through all known C<%client_ports> and relay the message to them. |
404 |
|
405 |
And thats it. |
406 |
|
407 |
=head2 The Remaining Problems |
408 |
|
409 |
The shown implementation still has some bugs. For instance: How does the |
410 |
server know that the client isn't there anymore, and can cleanup the |
411 |
C<%client_ports> hash? And also the chat messages have no originator, |
412 |
so we don't know who actually sent the message (which would be quite useful |
413 |
for human-to-human interaction: to know who the other one is :). |
414 |
|
415 |
But aside from these issues I hope this tutorial got you the swing of |
416 |
L<AnyEvent::MP> and explained some common idioms. |
417 |
|
418 |
How to solve the reliability and C<%client_ports> cleanup problem will |
419 |
be explained later in this tutorial (TODO). |
420 |
|
421 |
=head2 Inside The Protocol |
422 |
|
423 |
Now, for the interested parties, let me explain some details about the protocol |
424 |
that L<AnyEvent::MP> nodes use to communicate to each other. If you are not |
425 |
interested you can skip this section. |
426 |
|
427 |
Usually TCP is used for communication. Each I<node>, if configured to be a |
428 |
I<public> node with the C<become_public> function will listen on the configured |
429 |
TCP port (default is usually 4040). |
430 |
|
431 |
If now one I<node> wants to send a message to another I<node> it will connect |
432 |
to the host and port given in the I<port id>. |
433 |
|
434 |
Then some handshaking occurs to check whether both I<nodes> have the same |
435 |
I<shared secret> configured. Optionally even TLS can be enabled (about how to |
436 |
do this exactly please consult the L<AnyEvent::MP> man pages, just a hint: It |
437 |
should be enough to put the private key and (self signed) certificate in the |
438 |
C<~/.aemp-secret> file of all nodes). |
439 |
|
440 |
Now the serialized (usually L<JSON> is used for this, but it is also possible |
441 |
to use other serialization formats, like L<Storable>) messages are sent over |
442 |
the wire using a simple framing protocol. |
443 |
|
444 |
=head1 SEE ALSO |
445 |
|
446 |
L<AnyEvent> |
447 |
|
448 |
L<AnyEvent::Handle> |
449 |
|
450 |
L<AnyEvent::MP> |
451 |
|
452 |
=head1 AUTHOR |
453 |
|
454 |
Robin Redeker <elmex@ta-sa.org> |
455 |
|