1 |
root |
1.4 |
=head1 Message Passing for the Non-Blocked Mind |
2 |
elmex |
1.1 |
|
3 |
root |
1.8 |
=head1 Introduction and Terminology |
4 |
elmex |
1.1 |
|
5 |
root |
1.4 |
This is a tutorial about how to get the swing of the new L<AnyEvent::MP> |
6 |
root |
1.23 |
module, which allows programs to transparently pass messages within the |
7 |
|
|
process and to other processes on the same or a different host. |
8 |
elmex |
1.1 |
|
9 |
root |
1.23 |
What kind of messages? Basically a message here means a list of Perl |
10 |
root |
1.15 |
strings, numbers, hashes and arrays, anything that can be expressed as a |
11 |
root |
1.43 |
L<JSON> text (as JSON is the default serialiser in the protocol). Here are |
12 |
|
|
two examples: |
13 |
elmex |
1.1 |
|
14 |
root |
1.23 |
write_log => 1251555874, "action was successful.\n" |
15 |
|
|
123, ["a", "b", "c"], { foo => "bar" } |
16 |
elmex |
1.21 |
|
17 |
root |
1.23 |
When using L<AnyEvent::MP> it is customary to use a descriptive string as |
18 |
root |
1.46 |
first element of a message that indicates the type of the message. This |
19 |
root |
1.23 |
element is called a I<tag> in L<AnyEvent::MP>, as some API functions |
20 |
|
|
(C<rcv>) support matching it directly. |
21 |
|
|
|
22 |
|
|
Supposedly you want to send a ping message with your current time to |
23 |
|
|
somewhere, this is how such a message might look like (in Perl syntax): |
24 |
|
|
|
25 |
|
|
ping => 1251381636 |
26 |
|
|
|
27 |
|
|
Now that we know what a message is, to which entities are those |
28 |
|
|
messages being I<passed>? They are I<passed> to I<ports>. A I<port> is |
29 |
|
|
a destination for messages but also a context to execute code: when |
30 |
|
|
a runtime error occurs while executing code belonging to a port, the |
31 |
|
|
exception will be raised on the port and can even travel to interested |
32 |
|
|
parties on other nodes, which makes supervision of distributed processes |
33 |
|
|
easy. |
34 |
|
|
|
35 |
|
|
How do these ports relate to things you know? Each I<port> belongs |
36 |
|
|
to a I<node>, and a I<node> is just the UNIX process that runs your |
37 |
|
|
L<AnyEvent::MP> application. |
38 |
|
|
|
39 |
|
|
Each I<node> is distinguished from other I<nodes> running on the same or |
40 |
|
|
another host in a network by its I<node ID>. A I<node ID> is simply a |
41 |
|
|
unique string chosen manually or assigned by L<AnyEvent::MP> in some way |
42 |
|
|
(UNIX nodename, random string...). |
43 |
|
|
|
44 |
|
|
Here is a diagram about how I<nodes>, I<ports> and UNIX processes relate |
45 |
|
|
to each other. The setup consists of two nodes (more are of course |
46 |
|
|
possible): Node C<A> (in UNIX process 7066) with the ports C<ABC> and |
47 |
|
|
C<DEF>. And the node C<B> (in UNIX process 8321) with the ports C<FOO> and |
48 |
|
|
C<BAR>. |
49 |
elmex |
1.17 |
|
50 |
|
|
|
51 |
|
|
|- PID: 7066 -| |- PID: 8321 -| |
52 |
|
|
| | | | |
53 |
|
|
| Node ID: A | | Node ID: B | |
54 |
|
|
| | | | |
55 |
|
|
| Port ABC =|= <----\ /-----> =|= Port FOO | |
56 |
|
|
| | X | | |
57 |
|
|
| Port DEF =|= <----/ \-----> =|= Port BAR | |
58 |
|
|
| | | | |
59 |
|
|
|-------------| |-------------| |
60 |
|
|
|
61 |
root |
1.23 |
The strings for the I<port IDs> here are just for illustrative |
62 |
|
|
purposes: Even though I<ports> in L<AnyEvent::MP> are also identified by |
63 |
root |
1.43 |
strings, they can't be chosen manually and are assigned by the system |
64 |
root |
1.23 |
dynamically. These I<port IDs> are unique within a network and can also be |
65 |
root |
1.46 |
used to identify senders, or even as message tags for instance. |
66 |
root |
1.23 |
|
67 |
|
|
The next sections will explain the API of L<AnyEvent::MP> by going through |
68 |
|
|
a few simple examples. Later some more complex idioms are introduced, |
69 |
|
|
which are hopefully useful to solve some real world problems. |
70 |
root |
1.8 |
|
71 |
root |
1.39 |
=head2 Passing Your First Message |
72 |
elmex |
1.16 |
|
73 |
root |
1.46 |
For starters, let's have a look at the messaging API. The following |
74 |
|
|
example is just a demo to show the basic elements of message passing with |
75 |
root |
1.24 |
L<AnyEvent::MP>. |
76 |
|
|
|
77 |
|
|
The example should print: C<Ending with: 123>, in a rather complicated |
78 |
|
|
way, by passing some message to a port. |
79 |
elmex |
1.16 |
|
80 |
|
|
use AnyEvent; |
81 |
|
|
use AnyEvent::MP; |
82 |
|
|
|
83 |
|
|
my $end_cv = AnyEvent->condvar; |
84 |
|
|
|
85 |
|
|
my $port = port; |
86 |
|
|
|
87 |
|
|
rcv $port, test => sub { |
88 |
|
|
my ($data) = @_; |
89 |
|
|
$end_cv->send ($data); |
90 |
|
|
}; |
91 |
|
|
|
92 |
|
|
snd $port, test => 123; |
93 |
|
|
|
94 |
|
|
print "Ending with: " . $end_cv->recv . "\n"; |
95 |
|
|
|
96 |
root |
1.24 |
It already uses most of the essential functions inside |
97 |
root |
1.46 |
L<AnyEvent::MP>: First there is the C<port> function which creates a |
98 |
root |
1.24 |
I<port> and will return it's I<port ID>, a simple string. |
99 |
|
|
|
100 |
|
|
This I<port ID> can be used to send messages to the port and install |
101 |
|
|
handlers to receive messages on the port. Since it is a simple string |
102 |
|
|
it can be safely passed to other I<nodes> in the network when you want |
103 |
|
|
to refer to that specific port (usually used for RPC, where you need |
104 |
|
|
to tell the other end which I<port> to send the reply to - messages in |
105 |
|
|
L<AnyEvent::MP> have a destination, but no source). |
106 |
elmex |
1.17 |
|
107 |
root |
1.24 |
The next function is C<rcv>: |
108 |
elmex |
1.16 |
|
109 |
elmex |
1.17 |
rcv $port, test => sub { ... }; |
110 |
elmex |
1.16 |
|
111 |
root |
1.24 |
It installs a receiver callback on the I<port> that specified as the first |
112 |
|
|
argument (it only works for "local" ports, i.e. ports created on the same |
113 |
|
|
node). The next argument, in this example C<test>, specifies a I<tag> to |
114 |
|
|
match. This means that whenever a message with the first element being |
115 |
|
|
the string C<test> is received, the callback is called with the remaining |
116 |
elmex |
1.17 |
parts of that message. |
117 |
|
|
|
118 |
root |
1.24 |
Messages can be sent with the C<snd> function, which is used like this in |
119 |
|
|
the example above: |
120 |
elmex |
1.17 |
|
121 |
|
|
snd $port, test => 123; |
122 |
|
|
|
123 |
root |
1.24 |
This will send the message C<'test', 123> to the I<port> with the I<port |
124 |
|
|
ID> stored in C<$port>. Since in this case the receiver has a I<tag> match |
125 |
|
|
on C<test> it will call the callback with the first argument being the |
126 |
|
|
number C<123>. |
127 |
|
|
|
128 |
root |
1.43 |
The callback is a typical AnyEvent idiom: the callback just passes |
129 |
root |
1.24 |
that number on to the I<condition variable> C<$end_cv> which will then |
130 |
|
|
pass the value to the print. Condition variables are out of the scope |
131 |
|
|
of this tutorial and not often used with ports, so please consult the |
132 |
elmex |
1.17 |
L<AnyEvent::Intro> about them. |
133 |
|
|
|
134 |
root |
1.24 |
Passing messages inside just one process is boring. Before we can move on |
135 |
|
|
and do interprocess message passing we first have to make sure some things |
136 |
|
|
have been set up correctly for our nodes to talk to each other. |
137 |
elmex |
1.17 |
|
138 |
root |
1.39 |
=head2 System Requirements and System Setup |
139 |
elmex |
1.17 |
|
140 |
root |
1.25 |
Before we can start with real IPC we have to make sure some things work on |
141 |
|
|
your system. |
142 |
elmex |
1.17 |
|
143 |
root |
1.25 |
First we have to setup a I<shared secret>: for two L<AnyEvent::MP> |
144 |
|
|
I<nodes> to be able to communicate with each other over the network it is |
145 |
|
|
necessary to setup the same I<shared secret> for both of them, so they can |
146 |
|
|
prove their trustworthyness to each other. |
147 |
elmex |
1.17 |
|
148 |
|
|
The easiest way is to set this up is to use the F<aemp> utility: |
149 |
|
|
|
150 |
|
|
aemp gensecret |
151 |
|
|
|
152 |
root |
1.25 |
This creates a F<$HOME/.perl-anyevent-mp> config file and generates a |
153 |
|
|
random shared secret. You can copy this file to any other system and |
154 |
|
|
then communicate over the network (via TCP) with it. You can also select |
155 |
|
|
your own shared secret (F<aemp setsecret>) and for increased security |
156 |
|
|
requirements you can even create (or configure) a TLS certificate (F<aemp |
157 |
|
|
gencert>), causing connections to not just be securely authenticated, but |
158 |
|
|
also to be encrypted and protected against tinkering. |
159 |
|
|
|
160 |
|
|
Connections will only be successfully established when the I<nodes> |
161 |
|
|
that want to connect to each other have the same I<shared secret> (or |
162 |
|
|
successfully verify the TLS certificate of the other side, in which case |
163 |
|
|
no shared secret is required). |
164 |
elmex |
1.17 |
|
165 |
|
|
B<If something does not work as expected, and for example tcpdump shows |
166 |
|
|
that the connections are closed almost immediately, you should make sure |
167 |
|
|
that F<~/.perl-anyevent-mp> is the same on all hosts/user accounts that |
168 |
|
|
you try to connect with each other!> |
169 |
elmex |
1.16 |
|
170 |
root |
1.25 |
Thats is all for now, you will find some more advanced fiddling with the |
171 |
|
|
C<aemp> utility later. |
172 |
|
|
|
173 |
root |
1.35 |
=head2 Shooting the Trouble |
174 |
|
|
|
175 |
|
|
Sometimes things go wrong, and AnyEvent::MP, being a professional module, |
176 |
root |
1.43 |
does not gratuitously spill out messages to your screen. |
177 |
root |
1.35 |
|
178 |
|
|
To help troubleshooting any issues, there are two environment variables |
179 |
|
|
that you can set. The first, C<PERL_ANYEVENT_MP_WARNLEVEL> sets the |
180 |
|
|
logging level. The default is C<5>, which means nothing much is |
181 |
root |
1.43 |
printed. You can increase it to C<8> or C<9> to get more verbose |
182 |
root |
1.35 |
output. This is example output when starting a node: |
183 |
|
|
|
184 |
root |
1.46 |
2012-03-04 19:41:10 <8> node cerebro starting up. |
185 |
|
|
2012-03-04 19:41:10 <8> node listens on [10.0.0.1:4040]. |
186 |
|
|
2012-03-04 19:41:10 <9> trying connect to seed node 10.0.0.19:4040. |
187 |
|
|
2012-03-04 19:41:10 <9> 10.0.0.19:4040 connected as rain |
188 |
|
|
2012-03-04 19:41:10 <7> rain is up () |
189 |
root |
1.35 |
|
190 |
|
|
A lot of info, but at least you can see that it does something. |
191 |
|
|
|
192 |
|
|
The other environment variable that can be useful is |
193 |
|
|
C<PERL_ANYEVENT_MP_TRACE>, which, when set to a true value, will cause |
194 |
root |
1.46 |
most messages that are sent or received to be printed. For example, F<aemp |
195 |
|
|
restart rijk> might output these message exchanges: |
196 |
root |
1.35 |
|
197 |
root |
1.46 |
SND rijk <- [null,"eval","AnyEvent::Watchdog::Util::restart; ()","aemp/cerebro/z4kUPp2JT4#b"] |
198 |
|
|
SND rain <- [null,"g_slave",{"'l":{"aemp/cerebro/z4kUPp2JT4":["10.0.0.1:48168"]}}] |
199 |
|
|
SND rain <- [null,"g_find","rijk"] |
200 |
|
|
RCV rain -> ["","g_found","rijk",["10.0.0.23:4040"]] |
201 |
|
|
RCV rijk -> ["b",""] |
202 |
elmex |
1.18 |
|
203 |
root |
1.30 |
=head1 PART 1: Passing Messages Between Processes |
204 |
elmex |
1.18 |
|
205 |
|
|
=head2 The Receiver |
206 |
|
|
|
207 |
root |
1.25 |
Lets split the previous example up into two programs: one that contains |
208 |
|
|
the sender and one for the receiver. First the receiver application, in |
209 |
|
|
full: |
210 |
elmex |
1.18 |
|
211 |
|
|
use AnyEvent; |
212 |
|
|
use AnyEvent::MP; |
213 |
|
|
|
214 |
root |
1.45 |
configure nodeid => "eg_receiver/%u", binds => ["*:4040"]; |
215 |
elmex |
1.18 |
|
216 |
|
|
my $port = port; |
217 |
root |
1.47 |
db_set eg_receivers => $port; |
218 |
elmex |
1.18 |
|
219 |
|
|
rcv $port, test => sub { |
220 |
|
|
my ($data, $reply_port) = @_; |
221 |
|
|
|
222 |
|
|
print "Received data: " . $data . "\n"; |
223 |
|
|
}; |
224 |
|
|
|
225 |
|
|
AnyEvent->condvar->recv; |
226 |
|
|
|
227 |
|
|
=head3 AnyEvent::MP::Global |
228 |
|
|
|
229 |
root |
1.43 |
Now, that wasn't too bad, was it? OK, let's step through the new functions |
230 |
root |
1.47 |
that have been used. |
231 |
elmex |
1.18 |
|
232 |
root |
1.44 |
=head3 C<configure> and Joining and Maintaining the Network |
233 |
elmex |
1.18 |
|
234 |
root |
1.47 |
First let's have a look at C<configure>: |
235 |
elmex |
1.18 |
|
236 |
root |
1.47 |
configure nodeid => "eg_receiver/%u", binds => ["*:4040"]; |
237 |
elmex |
1.18 |
|
238 |
|
|
Before we are able to send messages to other nodes we have to initialise |
239 |
root |
1.26 |
ourself to become a "distributed node". Initialising a node means naming |
240 |
root |
1.47 |
the node and binding some TCP listeners so that other nodes can |
241 |
|
|
contact it. |
242 |
|
|
|
243 |
|
|
Additionally, to actually link all nodes in a network together, you can |
244 |
|
|
specify a number of seed addresses, which will be used by the node to |
245 |
|
|
connect itself into an existing network, as we will see shortly. |
246 |
root |
1.26 |
|
247 |
root |
1.28 |
All of this (and more) can be passed to the C<configure> function - later |
248 |
|
|
we will see how we can do all this without even passing anything to |
249 |
|
|
C<configure>! |
250 |
|
|
|
251 |
|
|
The first parameter, C<nodeid>, specified the node ID (in this case |
252 |
root |
1.47 |
C<eg_receiver/%u> - the default is to use the node name of the current |
253 |
|
|
host plus C</%u>, which goves the node a name with a random suffix to |
254 |
|
|
make it unique, but for this example we want the node to have a bit more |
255 |
|
|
personality, and name it C<eg_receiver> with a random suffix. |
256 |
|
|
|
257 |
|
|
Why the random suffix? Node IDs need to be unique within the network and |
258 |
|
|
appending a random suffix is the easiest way to do that. |
259 |
root |
1.28 |
|
260 |
|
|
The second parameter, C<binds>, specifies a list of C<address:port> pairs |
261 |
|
|
to bind TCP listeners on. The special "address" of C<*> means to bind on |
262 |
root |
1.47 |
every local IP address (this might not work on every OS, so explicit IP |
263 |
|
|
addresses are best). |
264 |
root |
1.28 |
|
265 |
|
|
The reason to bind on a TCP port is not just that other nodes can connect |
266 |
|
|
to us: if no binds are specified, the node will still bind on a dynamic |
267 |
|
|
port on all local addresses - but in this case we won't know the port, and |
268 |
|
|
cannot tell other nodes to connect to it as seed node. |
269 |
|
|
|
270 |
root |
1.47 |
Now, a I<seed> is simply the TCP address of some other node in the |
271 |
|
|
network, often the same string as used for the C<binds> parameter of the |
272 |
|
|
other node. The need for seeds is easy to explain: I<somehow> the nodes |
273 |
|
|
of an aemp network have to find each other, and often this means over the |
274 |
|
|
internet. So broadcasts are out. |
275 |
|
|
|
276 |
|
|
Instead, a node usually specifies the addresses of a few (for redundancy) |
277 |
|
|
other nodes, some of which should be up. Two nodes can set each other as |
278 |
|
|
seeds without any issues. You could even specify all nodes as seeds for |
279 |
|
|
all nodes, for total redundancy. But the common case is to have some more |
280 |
|
|
or less central, stable servers running seed services for other nodes. |
281 |
|
|
|
282 |
|
|
All you need to do to ensure that an AnyEvent::MP network connects |
283 |
|
|
together is to make sure that all connections from nodes to their seed |
284 |
|
|
nodes I<somehow> span the whole network. The simplest way to do that would |
285 |
|
|
be for all nodes to specify a single node as seed node, and you would get |
286 |
|
|
a star topology. If you specify all nodes as seed nodes, you get a fully |
287 |
|
|
meshed network (that's what previous releases of AnyEvent::MP actually |
288 |
|
|
did). |
289 |
|
|
|
290 |
|
|
A node tries to keep connections open to all of it's seed nodes at all |
291 |
|
|
times, while other connections are made on demand only. |
292 |
|
|
|
293 |
|
|
All of this ensures that the network stays one network - even if all the |
294 |
|
|
nodes in one half of the net are separated from the nodes in the other |
295 |
|
|
half by some network problem, once that is over, they will eventually |
296 |
|
|
become a single network again. |
297 |
|
|
|
298 |
|
|
In addition to creating the network, a node also expects the seed nodes to |
299 |
|
|
run the shared database service - if need be, by automatically starting it, |
300 |
|
|
so you don't normally need to configure this explicitly. |
301 |
|
|
|
302 |
|
|
#TODO# later?#d# |
303 |
|
|
The process of joining a network takes time, during which the node |
304 |
|
|
is already running. This means it takes time until the node is |
305 |
|
|
fully connected, and information about services in the network are |
306 |
|
|
available. This is why most AnyEvent::MP programs start by waiting a while |
307 |
|
|
until the information they need is available. |
308 |
|
|
|
309 |
|
|
We will see how this is done later, in the sender program. |
310 |
elmex |
1.19 |
|
311 |
root |
1.28 |
=head3 Registering the Receiver |
312 |
elmex |
1.19 |
|
313 |
root |
1.47 |
Coming back to our example, after the node has been configured for network |
314 |
|
|
access, it is time to publish some service, namely the receive service. |
315 |
elmex |
1.19 |
|
316 |
root |
1.47 |
For that, let's look at the next lines: |
317 |
elmex |
1.19 |
|
318 |
|
|
my $port = port; |
319 |
root |
1.47 |
db_set eg_receivers => $port; |
320 |
elmex |
1.19 |
|
321 |
root |
1.27 |
The C<port> function has already been discussed. It simply creates a new |
322 |
root |
1.47 |
I<port> and returns the I<port ID>. The C<db_reg> function, however, is |
323 |
|
|
new: The first argument is the name of a I<database family> and the second |
324 |
|
|
argument is the name of a I<subkey> within that family. The third argument |
325 |
|
|
would be the I<value> to be associated with the family and subkey, but, |
326 |
|
|
since it is missing, it will simply be C<undef>. |
327 |
|
|
|
328 |
|
|
Ok, what's this weird tlak about families you wonder - AnyEvent::MP comes |
329 |
|
|
with a distributed database. This database runs on so-called "global" |
330 |
|
|
nodes, which usually are the seed nodes of your network. The database |
331 |
|
|
structure is "simply" a hash of hashes of values. |
332 |
|
|
|
333 |
|
|
In other words, if the database were stored in C<%DB>, then the C<db_set> |
334 |
|
|
function more or less would do this: |
335 |
|
|
|
336 |
|
|
$DB{eg_receivers}{$port} = undef; |
337 |
|
|
|
338 |
|
|
So the ominous "family" selects a hash in the database, and the "subkey" |
339 |
|
|
is simply the key in this hash. And C<db_set> very much works like an |
340 |
|
|
assignment. |
341 |
|
|
|
342 |
|
|
The family namespace is shared by all nodes in a network, so the names |
343 |
|
|
should be reasonably unique, for example, they could start with the name |
344 |
|
|
of your module, or the name of the program. |
345 |
root |
1.27 |
|
346 |
root |
1.47 |
The purpose behind adding this key to the database is that the sender can |
347 |
|
|
look it up and find our port. We will shortly see how. |
348 |
root |
1.27 |
|
349 |
|
|
The last step in the example is to set up a receiver callback for those |
350 |
|
|
messages, just as was discussed in the first example. We again match |
351 |
|
|
for the tag C<test>. The difference is that this time we don't exit the |
352 |
|
|
application after receiving the first message. Instead we continue to wait |
353 |
|
|
for new messages indefinitely. |
354 |
elmex |
1.19 |
|
355 |
elmex |
1.20 |
=head2 The Sender |
356 |
root |
1.8 |
|
357 |
root |
1.27 |
Ok, now let's take a look at the sender code: |
358 |
root |
1.4 |
|
359 |
elmex |
1.1 |
use AnyEvent; |
360 |
|
|
use AnyEvent::MP; |
361 |
|
|
|
362 |
root |
1.45 |
configure nodeid => "eg_sender/%u", seeds => ["*:4040"]; |
363 |
elmex |
1.1 |
|
364 |
root |
1.47 |
my $guard = db_mon eg_receivers => sub { |
365 |
|
|
my ($family, $keys) = @_; |
366 |
|
|
return unless %$family; |
367 |
|
|
|
368 |
|
|
# now there are some receivers, send them a message |
369 |
|
|
snd $_ => test => time, keys %$family |
370 |
|
|
for keys %$family; |
371 |
|
|
}; |
372 |
elmex |
1.1 |
|
373 |
|
|
AnyEvent->condvar->recv; |
374 |
|
|
|
375 |
root |
1.28 |
It's even less code. The C<configure> serves the same purpose as in the |
376 |
|
|
receiver, but instead of specifying binds we specify a list of seeds - |
377 |
root |
1.47 |
the seed happens to be the same as the bind used by the receiver, which |
378 |
|
|
therefore becomes our seed node. |
379 |
root |
1.27 |
|
380 |
root |
1.47 |
Remember the part about having to wait till things become available? After |
381 |
|
|
configure returns, nothing has been done yet - the node is not connected |
382 |
|
|
to the network, knows nothing about the database contents, and it can take |
383 |
|
|
ages (for a computer :) for this situation to change. |
384 |
|
|
|
385 |
|
|
Therefore, the sender waits, in this case by using the C<db_mon> |
386 |
|
|
function. This function registers an interest in a specific database |
387 |
|
|
family (C<eg_receivers>). Each time something inside the family changes |
388 |
|
|
(a key is added, changed or deleted), it will call our callback with the |
389 |
|
|
family hash as first argument, and the list of keys as second argument. |
390 |
|
|
|
391 |
|
|
The callback only checks whether the C<%$family> has is empty - in this |
392 |
|
|
case it does nothing. |
393 |
|
|
|
394 |
|
|
Eventually the family will, however, contain the port we set in the |
395 |
|
|
sender. Then it will send a message to it and any other receiver in the |
396 |
|
|
group. |
397 |
|
|
|
398 |
|
|
You can experiment by having multiple receivers - you have to change the |
399 |
|
|
"binds" parameter in the receiver to the seeds used in the sender to start |
400 |
|
|
up additional receivers, but then you can start as many as you like. If |
401 |
|
|
you specify proper IP addresses for the seeds, you can even run them on |
402 |
|
|
different computers. |
403 |
|
|
|
404 |
|
|
Each time you start the sender, it will send a message to all receivers it |
405 |
|
|
finds (you have to interrupt it manualy afterwards). |
406 |
|
|
|
407 |
|
|
Things you could try include using C<PERL_ANYEVENT_MP_TRACE=1> to see |
408 |
|
|
which messages are exchanged, or starting the sender first and see how |
409 |
|
|
long it takes it to find the receiver. |
410 |
root |
1.27 |
|
411 |
root |
1.28 |
=head3 Splitting Network Configuration and Application Code |
412 |
|
|
|
413 |
root |
1.47 |
#TODO# |
414 |
root |
1.43 |
OK, so far, this works. In the real world, however, the person configuring |
415 |
root |
1.28 |
your application to run on a specific network (the end user or network |
416 |
|
|
administrator) is often different to the person coding the application. |
417 |
|
|
|
418 |
|
|
Or to put it differently: the arguments passed to configure are usually |
419 |
elmex |
1.31 |
provided not by the programmer, but by whoever is deploying the program. |
420 |
root |
1.28 |
|
421 |
|
|
To make this easy, AnyEvent::MP supports a simple configuration database, |
422 |
|
|
using profiles, which can be managed using the F<aemp> command-line |
423 |
root |
1.30 |
utility (yes, this section is about the advanced tinkering we mentioned |
424 |
|
|
before). |
425 |
root |
1.28 |
|
426 |
|
|
When you change both programs above to simply call |
427 |
|
|
|
428 |
|
|
configure; |
429 |
|
|
|
430 |
|
|
then AnyEvent::MP tries to look up a profile using the current node name |
431 |
|
|
in its configuration database, falling back to some global default. |
432 |
|
|
|
433 |
|
|
You can run "generic" nodes using the F<aemp> utility as well, and we will |
434 |
|
|
exploit this in the following way: we configure a profile "seed" and run |
435 |
|
|
a node using it, whose sole purpose is to be a seed node for our example |
436 |
|
|
programs. |
437 |
|
|
|
438 |
|
|
We bind the seed node to port 4040 on all interfaces: |
439 |
|
|
|
440 |
root |
1.29 |
aemp profile seed binds "*:4040" |
441 |
root |
1.28 |
|
442 |
|
|
And we configure all nodes to use this as seed node (this only works when |
443 |
|
|
running on the same host, for multiple machines you would provide the IP |
444 |
root |
1.30 |
address or hostname of the node running the seed), and use a random name |
445 |
|
|
(because we want to start multiple nodes on the same host): |
446 |
root |
1.28 |
|
447 |
root |
1.30 |
aemp seeds "*:4040" nodeid anon/ |
448 |
root |
1.28 |
|
449 |
|
|
Then we run the seed node: |
450 |
|
|
|
451 |
|
|
aemp run profile seed |
452 |
|
|
|
453 |
|
|
After that, we can start as many other nodes as we want, and they will all |
454 |
|
|
use our generic seed node to discover each other. |
455 |
root |
1.27 |
|
456 |
root |
1.28 |
In fact, starting many receivers nicely illustrates that the time sender |
457 |
|
|
can have multiple receivers. |
458 |
elmex |
1.7 |
|
459 |
root |
1.30 |
That's all for now - next we will teach you about monitoring by writing a |
460 |
|
|
simple chat client and server :) |
461 |
|
|
|
462 |
|
|
=head1 PART 2: Monitoring, Supervising, Exception Handling and Recovery |
463 |
|
|
|
464 |
|
|
That's a mouthful, so what does it mean? Our previous example is what one |
465 |
|
|
could call "very loosely coupled" - the sender doesn't care about whether |
466 |
|
|
there are any receivers, and the receivers do not care if there is any |
467 |
|
|
sender. |
468 |
|
|
|
469 |
|
|
This can work fine for simple services, but most real-world applications |
470 |
|
|
want to ensure that the side they are expecting to be there is actually |
471 |
|
|
there. Going one step further: most bigger real-world applications even |
472 |
|
|
want to ensure that if some component is missing, or has crashed, it will |
473 |
|
|
still be there, by recovering and restarting the service. |
474 |
|
|
|
475 |
|
|
AnyEvent::MP supports this by catching exceptions and network problems, |
476 |
|
|
and notifying interested parties of this. |
477 |
|
|
|
478 |
root |
1.41 |
=head2 Exceptions, Port Context, Network Errors and Monitors |
479 |
root |
1.30 |
|
480 |
|
|
=head3 Exceptions |
481 |
|
|
|
482 |
|
|
Exceptions are handled on a per-port basis: receive callbacks are executed |
483 |
root |
1.41 |
in a special context, the so-called I<port-context>: code that throws an |
484 |
|
|
otherwise uncaught exception will cause the port to be C<kil>led. Killed |
485 |
|
|
ports are destroyed automatically (killing ports is the only way to free |
486 |
|
|
ports, incidentally). |
487 |
root |
1.30 |
|
488 |
|
|
Ports can be monitored, even from a different host, and when a port is |
489 |
|
|
killed any entity monitoring it will be notified. |
490 |
|
|
|
491 |
|
|
Here is a simple example: |
492 |
|
|
|
493 |
|
|
use AnyEvent::MP; |
494 |
|
|
|
495 |
|
|
# create a port, it always dies |
496 |
|
|
my $port = port { die "oops" }; |
497 |
|
|
|
498 |
|
|
# monitor it |
499 |
|
|
mon $port, sub { |
500 |
|
|
warn "$port was killed (with reason @_)"; |
501 |
|
|
}; |
502 |
|
|
|
503 |
|
|
# now send it some message, causing it to die: |
504 |
|
|
snd $port; |
505 |
|
|
|
506 |
|
|
It first creates a port whose only action is to throw an exception, |
507 |
|
|
and the monitors it with the C<mon> function. Afterwards it sends it a |
508 |
|
|
message, causing it to die and call the monitoring callback: |
509 |
|
|
|
510 |
|
|
anon/6WmIpj.a was killed (with reason die oops at xxx line 5.) at xxx line 9. |
511 |
|
|
|
512 |
|
|
The callback was actually passed two arguments: C<die> (to indicate it did |
513 |
|
|
throw an exception as opposed to, say, a network error) and the exception |
514 |
|
|
message itself. |
515 |
|
|
|
516 |
|
|
What happens when a port is killed before we have a chance to monitor |
517 |
|
|
it? Granted, this is highly unlikely in our example, but when you program |
518 |
|
|
in a network this can easily happen due to races between nodes. |
519 |
|
|
|
520 |
|
|
use AnyEvent::MP; |
521 |
|
|
|
522 |
|
|
my $port = port { die "oops" }; |
523 |
|
|
|
524 |
|
|
snd $port; |
525 |
|
|
|
526 |
|
|
mon $port, sub { |
527 |
|
|
warn "$port was killed (with reason @_)"; |
528 |
|
|
}; |
529 |
|
|
|
530 |
|
|
This time we will get something like: |
531 |
|
|
|
532 |
|
|
anon/zpX.a was killed (with reason no_such_port cannot monitor nonexistent port) |
533 |
|
|
|
534 |
|
|
Since the port was already gone, the kill reason is now C<no_such_port> |
535 |
|
|
with some descriptive (we hope) error message. |
536 |
|
|
|
537 |
|
|
In fact, the kill reason is usually some identifier as first argument |
538 |
|
|
and a human-readable error message as second argument, but can be about |
539 |
|
|
anything (it's a list) or even nothing - which is called a "normal" kill. |
540 |
|
|
|
541 |
|
|
You can kill ports manually using the C<kil> function, which will be |
542 |
|
|
treated like an error when any reason is specified: |
543 |
|
|
|
544 |
|
|
kil $port, custom_error => "don't like your steenking face"; |
545 |
|
|
|
546 |
|
|
And a clean kill without any reason arguments: |
547 |
|
|
|
548 |
|
|
kil $port; |
549 |
|
|
|
550 |
|
|
By now you probably wonder what this "normal" kill business is: A common |
551 |
|
|
idiom is to not specify a callback to C<mon>, but another port, such as |
552 |
|
|
C<$SELF>: |
553 |
|
|
|
554 |
|
|
mon $port, $SELF; |
555 |
|
|
|
556 |
|
|
This basically means "monitor $port and kill me when it crashes". And a |
557 |
|
|
"normal" kill does not count as a crash. This way you can easily link |
558 |
|
|
ports together and make them crash together on errors (but allow you to |
559 |
|
|
remove a port silently). |
560 |
|
|
|
561 |
root |
1.34 |
=head3 Port Context |
562 |
|
|
|
563 |
|
|
When code runs in an environment where C<$SELF> contains its own port ID |
564 |
|
|
and exceptions will be caught, it is said to run in a port context. |
565 |
|
|
|
566 |
|
|
Since AnyEvent::MP is event-based, it is not uncommon to register |
567 |
|
|
callbacks from C<rcv> handlers. As example, assume that the port receive |
568 |
|
|
handler wants to C<die> a second later, using C<after>: |
569 |
|
|
|
570 |
|
|
my $port = port { |
571 |
|
|
after 1, sub { die "oops" }; |
572 |
|
|
}; |
573 |
|
|
|
574 |
|
|
Then you will find it does not work - when the after callback is executed, |
575 |
|
|
it does not run in port context anymore, so exceptions will not be caught. |
576 |
|
|
|
577 |
root |
1.41 |
For these cases, AnyEvent::MP exports a special "closure constructor" |
578 |
root |
1.43 |
called C<psub>, which works just like perls built-in C<sub>: |
579 |
root |
1.34 |
|
580 |
|
|
my $port = port { |
581 |
|
|
after 1, psub { die "oops" }; |
582 |
|
|
}; |
583 |
|
|
|
584 |
|
|
C<psub> stores C<$SELF> and returns a code reference. When the code |
585 |
|
|
reference is invoked, it will run the code block within the context of |
586 |
|
|
that port, so exception handling once more works as expected. |
587 |
|
|
|
588 |
root |
1.41 |
There is also a way to temporarily execute code in the context of some |
589 |
|
|
port, namely C<peval>: |
590 |
|
|
|
591 |
|
|
peval $port, sub { |
592 |
|
|
# die'ing here will kil $port |
593 |
|
|
}; |
594 |
|
|
|
595 |
|
|
The C<peval> function temporarily replaces C<$SELF> by the given C<$port> |
596 |
|
|
and then executes the given sub in a port context. |
597 |
|
|
|
598 |
root |
1.30 |
=head3 Network Errors and the AEMP Guarantee |
599 |
|
|
|
600 |
|
|
I mentioned another important source of monitoring failures: network |
601 |
|
|
problems. When a node loses connection to another node, it will invoke all |
602 |
|
|
monitoring actions as if the port was killed, even if it is possible that |
603 |
elmex |
1.31 |
the port still lives happily on another node (not being able to talk to a |
604 |
root |
1.30 |
node means we have no clue what's going on with it, it could be crashed, |
605 |
|
|
but also still running without knowing we lost the connection). |
606 |
|
|
|
607 |
|
|
So another way to view monitors is "notify me when some of my messages |
608 |
|
|
couldn't be delivered". AEMP has a guarantee about message delivery to a |
609 |
|
|
port: After starting a monitor, any message sent to a port will either |
610 |
|
|
be delivered, or, when it is lost, any further messages will also be lost |
611 |
elmex |
1.31 |
until the monitoring action is invoked. After that, further messages |
612 |
root |
1.30 |
I<might> get delivered again. |
613 |
|
|
|
614 |
|
|
This doesn't sound like a very big guarantee, but it is kind of the best |
615 |
elmex |
1.31 |
you can get while staying sane: Specifically, it means that there will |
616 |
|
|
be no "holes" in the message sequence: all messages sent are delivered |
617 |
root |
1.30 |
in order, without any missing in between, and when some were lost, you |
618 |
|
|
I<will> be notified of that, so you can take recovery action. |
619 |
|
|
|
620 |
|
|
=head3 Supervising |
621 |
|
|
|
622 |
|
|
Ok, so what is this crashing-everything-stuff going to make applications |
623 |
|
|
I<more> stable? Well in fact, the goal is not really to make them more |
624 |
|
|
stable, but to make them more resilient against actual errors and |
625 |
|
|
crashes. And this is not done by crashing I<everything>, but by crashing |
626 |
|
|
everything except a supervisor. |
627 |
|
|
|
628 |
elmex |
1.31 |
A supervisor is simply some code that ensures that an application (or a |
629 |
root |
1.30 |
part of it) is running, and if it crashes, is restarted properly. |
630 |
|
|
|
631 |
|
|
To show how to do all this we will create a simple chat server that can |
632 |
|
|
handle many chat clients. Both server and clients can be killed and |
633 |
|
|
restarted, and even crash, to some extent. |
634 |
|
|
|
635 |
|
|
=head2 Chatting, the Resilient Way |
636 |
|
|
|
637 |
|
|
Without further ado, here is the chat server (to run it, we assume the |
638 |
|
|
set-up explained earlier, with a separate F<aemp run> seed node): |
639 |
|
|
|
640 |
|
|
use common::sense; |
641 |
|
|
use AnyEvent::MP; |
642 |
|
|
use AnyEvent::MP::Global; |
643 |
|
|
|
644 |
|
|
configure; |
645 |
|
|
|
646 |
|
|
my %clients; |
647 |
|
|
|
648 |
|
|
sub msg { |
649 |
|
|
print "relaying: $_[0]\n"; |
650 |
|
|
snd $_, $_[0] |
651 |
|
|
for values %clients; |
652 |
|
|
} |
653 |
|
|
|
654 |
|
|
our $server = port; |
655 |
|
|
|
656 |
|
|
rcv $server, join => sub { |
657 |
|
|
my ($client, $nick) = @_; |
658 |
|
|
|
659 |
|
|
$clients{$client} = $client; |
660 |
|
|
|
661 |
|
|
mon $client, sub { |
662 |
|
|
delete $clients{$client}; |
663 |
|
|
msg "$nick (quits, @_)"; |
664 |
|
|
}; |
665 |
|
|
msg "$nick (joins)"; |
666 |
|
|
}; |
667 |
|
|
|
668 |
|
|
rcv $server, privmsg => sub { |
669 |
|
|
my ($nick, $msg) = @_; |
670 |
|
|
msg "$nick: $msg"; |
671 |
|
|
}; |
672 |
|
|
|
673 |
root |
1.40 |
grp_reg eg_chat_server => $server; |
674 |
root |
1.30 |
|
675 |
|
|
warn "server ready.\n"; |
676 |
|
|
|
677 |
|
|
AnyEvent->condvar->recv; |
678 |
|
|
|
679 |
elmex |
1.31 |
Looks like a lot, but it is actually quite simple: after your usual |
680 |
root |
1.30 |
preamble (this time we use common sense), we define a helper function that |
681 |
|
|
sends some message to every registered chat client: |
682 |
|
|
|
683 |
|
|
sub msg { |
684 |
|
|
print "relaying: $_[0]\n"; |
685 |
|
|
snd $_, $_[0] |
686 |
|
|
for values %clients; |
687 |
|
|
} |
688 |
|
|
|
689 |
|
|
The clients are stored in the hash C<%client>. Then we define a server |
690 |
|
|
port and install two receivers on it, C<join>, which is sent by clients |
691 |
|
|
to join the chat, and C<privmsg>, that clients use to send actual chat |
692 |
|
|
messages. |
693 |
|
|
|
694 |
|
|
C<join> is most complicated. It expects the client port and the nickname |
695 |
|
|
to be passed in the message, and registers the client in C<%clients>. |
696 |
|
|
|
697 |
|
|
rcv $server, join => sub { |
698 |
|
|
my ($client, $nick) = @_; |
699 |
|
|
|
700 |
|
|
$clients{$client} = $client; |
701 |
|
|
|
702 |
|
|
The next step is to monitor the client. The monitoring action removes the |
703 |
|
|
client and sends a quit message with the error to all remaining clients. |
704 |
|
|
|
705 |
|
|
mon $client, sub { |
706 |
|
|
delete $clients{$client}; |
707 |
|
|
msg "$nick (quits, @_)"; |
708 |
|
|
}; |
709 |
|
|
|
710 |
|
|
And finally, it creates a join message and sends it to all clients. |
711 |
|
|
|
712 |
|
|
msg "$nick (joins)"; |
713 |
|
|
}; |
714 |
|
|
|
715 |
|
|
The C<privmsg> callback simply broadcasts the message to all clients: |
716 |
|
|
|
717 |
|
|
rcv $server, privmsg => sub { |
718 |
|
|
my ($nick, $msg) = @_; |
719 |
|
|
msg "$nick: $msg"; |
720 |
|
|
}; |
721 |
|
|
|
722 |
elmex |
1.31 |
And finally, the server registers itself in the server group, so that |
723 |
root |
1.30 |
clients can find it: |
724 |
|
|
|
725 |
root |
1.40 |
grp_reg eg_chat_server => $server; |
726 |
root |
1.30 |
|
727 |
|
|
Well, well... and where is this supervisor stuff? Well... we cheated, |
728 |
|
|
it's not there. To not overcomplicate the example, we only put it into |
729 |
|
|
the..... CLIENT! |
730 |
|
|
|
731 |
|
|
=head3 The Client, and a Supervisor! |
732 |
|
|
|
733 |
|
|
Again, here is the client, including supervisor, which makes it a bit |
734 |
|
|
longer: |
735 |
|
|
|
736 |
|
|
use common::sense; |
737 |
|
|
use AnyEvent::MP; |
738 |
|
|
use AnyEvent::MP::Global; |
739 |
|
|
|
740 |
|
|
my $nick = shift; |
741 |
|
|
|
742 |
|
|
configure; |
743 |
|
|
|
744 |
|
|
my ($client, $server); |
745 |
|
|
|
746 |
|
|
sub server_connect { |
747 |
root |
1.40 |
my $servernodes = grp_get "eg_chat_server" |
748 |
root |
1.30 |
or return after 1, \&server_connect; |
749 |
|
|
|
750 |
|
|
print "\rconnecting...\n"; |
751 |
|
|
|
752 |
|
|
$client = port { print "\r \r@_\n> " }; |
753 |
|
|
mon $client, sub { |
754 |
|
|
print "\rdisconnected @_\n"; |
755 |
|
|
&server_connect; |
756 |
|
|
}; |
757 |
|
|
|
758 |
|
|
$server = $servernodes->[0]; |
759 |
|
|
snd $server, join => $client, $nick; |
760 |
|
|
mon $server, $client; |
761 |
|
|
} |
762 |
|
|
|
763 |
|
|
server_connect; |
764 |
|
|
|
765 |
root |
1.34 |
my $w = AnyEvent->io (fh => 0, poll => 'r', cb => sub { |
766 |
root |
1.30 |
chomp (my $line = <STDIN>); |
767 |
|
|
print "> "; |
768 |
|
|
snd $server, privmsg => $nick, $line |
769 |
|
|
if $server; |
770 |
|
|
}); |
771 |
|
|
|
772 |
|
|
$| = 1; |
773 |
|
|
print "> "; |
774 |
|
|
AnyEvent->condvar->recv; |
775 |
|
|
|
776 |
|
|
The first thing the client does is to store the nick name (which is |
777 |
|
|
expected as the only command line argument) in C<$nick>, for further |
778 |
|
|
usage. |
779 |
|
|
|
780 |
|
|
The next relevant thing is... finally... the supervisor: |
781 |
|
|
|
782 |
|
|
sub server_connect { |
783 |
root |
1.40 |
my $servernodes = grp_get "eg_chat_server" |
784 |
root |
1.30 |
or return after 1, \&server_connect; |
785 |
|
|
|
786 |
|
|
This looks up the server in the C<eg_chat_server> global group. If it |
787 |
|
|
cannot find it (which is likely when the node is just starting up), |
788 |
|
|
it will wait a second and then retry. This "wait a bit and retry" |
789 |
|
|
is an important pattern, as distributed programming means lots of |
790 |
|
|
things are going on asynchronously. In practise, one should use a more |
791 |
|
|
intelligent algorithm, to possibly warn after an excessive number of |
792 |
|
|
retries. Hopefully future versions of AnyEvent::MP will offer some |
793 |
|
|
predefined supervisors, for now you will have to code it on your own. |
794 |
|
|
|
795 |
|
|
Next it creates a local port for the server to send messages to, and |
796 |
|
|
monitors it. When the port is killed, it will print "disconnected" and |
797 |
|
|
tell the supervisor function to retry again. |
798 |
|
|
|
799 |
|
|
$client = port { print "\r \r@_\n> " }; |
800 |
|
|
mon $client, sub { |
801 |
|
|
print "\rdisconnected @_\n"; |
802 |
|
|
&server_connect; |
803 |
|
|
}; |
804 |
|
|
|
805 |
|
|
Then everything is ready: the client will send a C<join> message with it's |
806 |
|
|
local port to the server, and start monitoring it: |
807 |
|
|
|
808 |
|
|
$server = $servernodes->[0]; |
809 |
|
|
snd $server, join => $client, $nick; |
810 |
|
|
mon $server, $client; |
811 |
|
|
} |
812 |
|
|
|
813 |
|
|
The monitor will ensure that if the server crashes or goes away, the |
814 |
|
|
client will be killed as well. This tells the user that the client was |
815 |
|
|
disconnected, and will then start to connect the server again. |
816 |
|
|
|
817 |
|
|
The rest of the program deals with the boring details of actually invoking |
818 |
|
|
the supervisor function to start the whole client process and handle the |
819 |
|
|
actual terminal input, sending it to the server. |
820 |
|
|
|
821 |
elmex |
1.31 |
You should now try to start the server and one or more clients in different |
822 |
root |
1.30 |
terminal windows (and the seed node): |
823 |
|
|
|
824 |
|
|
perl eg/chat_client nick1 |
825 |
|
|
perl eg/chat_client nick2 |
826 |
|
|
perl eg/chat_server |
827 |
|
|
aemp run profile seed |
828 |
|
|
|
829 |
|
|
And then you can experiment with chatting, killing one or more clients, or |
830 |
|
|
stopping and restarting the server, to see the monitoring in action. |
831 |
|
|
|
832 |
root |
1.33 |
The crucial point you should understand from this example is that |
833 |
|
|
monitoring is usually symmetric: when you monitor some other port, |
834 |
|
|
potentially on another node, that other port usually should monitor you, |
835 |
|
|
too, so when the connection dies, both ports get killed, or at least both |
836 |
|
|
sides can take corrective action. Exceptions are "servers" that serve |
837 |
|
|
multiple clients at once and might only wish to clean up, and supervisors, |
838 |
|
|
who of course should not normally get killed (unless they, too, have a |
839 |
|
|
supervisor). |
840 |
|
|
|
841 |
|
|
If you often think in object-oriented terms, then treat a port as an |
842 |
|
|
object, C<port> is the constructor, the receive callbacks set by C<rcv> |
843 |
|
|
act as methods, the C<kil> function becomes the explicit destructor and |
844 |
|
|
C<mon> installs a destructor hook. Unlike conventional object oriented |
845 |
|
|
programming, it can make sense to exchange ports more freely (for example, |
846 |
|
|
to monitor one port from another). |
847 |
|
|
|
848 |
root |
1.30 |
There is ample room for improvement: the server should probably remember |
849 |
|
|
the nickname in the C<join> handler instead of expecting it in every chat |
850 |
|
|
message, it should probably monitor itself, and the client should not try |
851 |
|
|
to send any messages unless a server is actually connected. |
852 |
|
|
|
853 |
|
|
=head1 PART 3: TIMTOWTDI: Virtual Connections |
854 |
|
|
|
855 |
root |
1.34 |
The chat system developed in the previous sections is very "traditional" |
856 |
|
|
in a way: you start some server(s) and some clients statically and they |
857 |
|
|
start talking to each other. |
858 |
|
|
|
859 |
|
|
Sometimes applications work more like "services": They can run on almost |
860 |
|
|
any node and talks to itself on other nodes. The L<AnyEvent::MP::Global> |
861 |
|
|
service for example monitors nodes joining the network and starts itself |
862 |
|
|
automatically on other nodes (if it isn't running already). |
863 |
|
|
|
864 |
|
|
A good way to design such applications is to put them into a module and |
865 |
|
|
create "virtual connections" to other nodes - we call this the "bridge |
866 |
|
|
head" method, because you start by creating a remote port (the bridge |
867 |
|
|
head) and from that you start to bootstrap your application. |
868 |
|
|
|
869 |
|
|
Since that sounds rather theoretical, let's redesign the chat server and |
870 |
|
|
client using this design method. |
871 |
|
|
|
872 |
|
|
Here is the server: |
873 |
|
|
|
874 |
|
|
use common::sense; |
875 |
|
|
use AnyEvent::MP; |
876 |
|
|
use AnyEvent::MP::Global; |
877 |
|
|
|
878 |
|
|
configure; |
879 |
|
|
|
880 |
root |
1.40 |
grp_reg eg_chat_server2 => $NODE; |
881 |
root |
1.34 |
|
882 |
|
|
my %clients; |
883 |
|
|
|
884 |
|
|
sub msg { |
885 |
|
|
print "relaying: $_[0]\n"; |
886 |
|
|
snd $_, $_[0] |
887 |
|
|
for values %clients; |
888 |
|
|
} |
889 |
|
|
|
890 |
|
|
sub client_connect { |
891 |
|
|
my ($client, $nick) = @_; |
892 |
|
|
|
893 |
|
|
mon $client; |
894 |
|
|
mon $client, sub { |
895 |
|
|
delete $clients{$client}; |
896 |
|
|
msg "$nick (quits, @_)"; |
897 |
|
|
}; |
898 |
|
|
|
899 |
|
|
$clients{$client} = $client; |
900 |
|
|
|
901 |
|
|
msg "$nick (joins)"; |
902 |
|
|
|
903 |
|
|
rcv $SELF, sub { msg "$nick: $_[0]" }; |
904 |
|
|
} |
905 |
|
|
|
906 |
|
|
warn "server ready.\n"; |
907 |
|
|
|
908 |
|
|
AnyEvent->condvar->recv; |
909 |
|
|
|
910 |
root |
1.39 |
It starts out not much different then the previous example, except that |
911 |
|
|
this time, we register the node port in the global group and not any port |
912 |
|
|
we created - the clients only want to know which node the server should be |
913 |
|
|
running on. In fact, they could also use some kind of election mechanism, |
914 |
|
|
to find the node with lowest load or something like that. |
915 |
|
|
|
916 |
|
|
The more interesting change is that indeed no server port is created - |
917 |
|
|
the server consists only of code, and "does" nothing by itself. All it |
918 |
|
|
does is define a function C<client_connect>, which expects a client port |
919 |
|
|
and a nick name as arguments. It then monitors the client port and binds |
920 |
|
|
a receive callback on C<$SELF>, which expects messages that in turn are |
921 |
|
|
broadcast to all clients. |
922 |
root |
1.34 |
|
923 |
|
|
The two C<mon> calls are a bit tricky - the first C<mon> is a shorthand |
924 |
|
|
for C<mon $client, $SELF>. The second does the normal "client has gone |
925 |
|
|
away" clean-up action. Both could actually be rolled into one C<mon> |
926 |
|
|
action. |
927 |
|
|
|
928 |
root |
1.39 |
C<$SELF> is a good hint that something interesting is going on. And |
929 |
|
|
indeed, when looking at the client code, there is a new function, |
930 |
|
|
C<spawn>: |
931 |
root |
1.34 |
|
932 |
|
|
use common::sense; |
933 |
|
|
use AnyEvent::MP; |
934 |
|
|
use AnyEvent::MP::Global; |
935 |
|
|
|
936 |
|
|
my $nick = shift; |
937 |
|
|
|
938 |
|
|
configure; |
939 |
|
|
|
940 |
|
|
$| = 1; |
941 |
|
|
|
942 |
|
|
my $port = port; |
943 |
|
|
|
944 |
|
|
my ($client, $server); |
945 |
|
|
|
946 |
|
|
sub server_connect { |
947 |
root |
1.40 |
my $servernodes = grp_get "eg_chat_server2" |
948 |
root |
1.34 |
or return after 1, \&server_connect; |
949 |
|
|
|
950 |
|
|
print "\rconnecting...\n"; |
951 |
|
|
|
952 |
|
|
$client = port { print "\r \r@_\n> " }; |
953 |
|
|
mon $client, sub { |
954 |
|
|
print "\rdisconnected @_\n"; |
955 |
|
|
&server_connect; |
956 |
|
|
}; |
957 |
|
|
|
958 |
|
|
$server = spawn $servernodes->[0], "::client_connect", $client, $nick; |
959 |
|
|
mon $server, $client; |
960 |
|
|
} |
961 |
|
|
|
962 |
|
|
server_connect; |
963 |
|
|
|
964 |
|
|
my $w = AnyEvent->io (fh => 0, poll => 'r', cb => sub { |
965 |
|
|
chomp (my $line = <STDIN>); |
966 |
|
|
print "> "; |
967 |
|
|
snd $server, $line |
968 |
|
|
if $server; |
969 |
|
|
}); |
970 |
|
|
|
971 |
|
|
print "> "; |
972 |
|
|
AnyEvent->condvar->recv; |
973 |
|
|
|
974 |
|
|
The client is quite similar to the previous one, but instead of contacting |
975 |
root |
1.39 |
the server I<port> (which no longer exists), it C<spawn>s (creates) a new |
976 |
|
|
the server I<port on node>: |
977 |
root |
1.34 |
|
978 |
|
|
$server = spawn $servernodes->[0], "::client_connect", $client, $nick; |
979 |
|
|
mon $server, $client; |
980 |
|
|
|
981 |
root |
1.39 |
And of course the first thing after creating it is monitoring it. |
982 |
root |
1.34 |
|
983 |
root |
1.39 |
The C<spawn> function creates a new port on a remote node and returns |
984 |
|
|
its port ID. After creating the port it calls a function on the remote |
985 |
|
|
node, passing any remaining arguments to it, and - most importantly - |
986 |
|
|
executes the function within the context of the new port, so it can be |
987 |
root |
1.43 |
manipulated by referring to C<$SELF>. The init function can reside in a |
988 |
root |
1.39 |
module (actually it normally I<should> reside in a module) - AnyEvent::MP |
989 |
|
|
will automatically load the module if the function isn't defined. |
990 |
|
|
|
991 |
|
|
The C<spawn> function returns immediately, which means you can instantly |
992 |
root |
1.34 |
send messages to the port, long before the remote node has even heard |
993 |
|
|
of our request to create a port on it. In fact, the remote node might |
994 |
|
|
not even be running. Despite these troubling facts, everything should |
995 |
|
|
work just fine: if the node isn't running (or the init function throws an |
996 |
|
|
exception), then the monitor will trigger because the port doesn't exist. |
997 |
|
|
|
998 |
|
|
If the spawn message gets delivered, but the monitoring message is not |
999 |
root |
1.39 |
because of network problems (extremely unlikely, but monitoring, after |
1000 |
|
|
all, is implemented by passing a message, and messages can get lost), then |
1001 |
|
|
this connection loss will eventually trigger the monitoring action. On the |
1002 |
|
|
remote node (which in return monitors the client) the port will also be |
1003 |
|
|
cleaned up on connection loss. When the remote node comes up again and our |
1004 |
|
|
monitoring message can be delivered, it will instantly fail because the |
1005 |
|
|
port has been cleaned up in the meantime. |
1006 |
root |
1.34 |
|
1007 |
|
|
If your head is spinning by now, that's fine - just keep in mind, after |
1008 |
root |
1.39 |
creating a port, monitor it on the local node, and monitor "the other |
1009 |
|
|
side" from the remote node, and all will be cleaned up just fine. |
1010 |
root |
1.34 |
|
1011 |
root |
1.36 |
=head2 Services |
1012 |
root |
1.34 |
|
1013 |
root |
1.36 |
Above it was mentioned that C<spawn> automatically loads modules, and this |
1014 |
|
|
can be exploited in various ways. |
1015 |
|
|
|
1016 |
|
|
Assume for a moment you put the server into a file called |
1017 |
|
|
F<mymod/chatserver.pm> reachable from the current directory. Then you |
1018 |
|
|
could run a node there with: |
1019 |
|
|
|
1020 |
|
|
aemp run |
1021 |
|
|
|
1022 |
|
|
The other nodes could C<spawn> the server by using |
1023 |
|
|
C<mymod::chatserver::client_connect> as init function. |
1024 |
|
|
|
1025 |
|
|
Likewise, when you have some service that starts automatically (similar to |
1026 |
|
|
AnyEvent::MP::Global), then you can configure this service statically: |
1027 |
|
|
|
1028 |
|
|
aemp profile mysrvnode services mymod::service:: |
1029 |
|
|
aemp run profile mysrvnode |
1030 |
|
|
|
1031 |
root |
1.39 |
And the module will automatically be loaded in the node, as specifying a |
1032 |
root |
1.38 |
module name (with C<::>-suffix) will simply load the module, which is then |
1033 |
|
|
free to do whatever it wants. |
1034 |
root |
1.36 |
|
1035 |
|
|
Of course, you can also do it in the much more standard way by writing |
1036 |
|
|
a module (e.g. C<BK::Backend::IRC>), installing it as part of a module |
1037 |
|
|
distribution and then configure nodes, for example, if I want to run the |
1038 |
|
|
Bummskraut IRC backend on a machine named "ruth", I could do this: |
1039 |
|
|
|
1040 |
|
|
aemp profile ruth addservice BK::Backend::IRC:: |
1041 |
|
|
|
1042 |
root |
1.43 |
And any F<aemp run> on that host will automatically have the Bummskraut |
1043 |
|
|
IRC backend running. |
1044 |
root |
1.36 |
|
1045 |
|
|
That's plenty of possibilities you can use - it's all up to you how you |
1046 |
|
|
structure your application. |
1047 |
elmex |
1.7 |
|
1048 |
root |
1.42 |
=head1 PART 4: Coro::MP - selective receive |
1049 |
|
|
|
1050 |
|
|
Not all problems lend themselves naturally to an event-based solution: |
1051 |
|
|
sometimes things are easier if you can decide in what order you want to |
1052 |
|
|
receive messages, irregardless of the order in which they were sent. |
1053 |
|
|
|
1054 |
|
|
In these cases, L<Coro::MP> can provide a nice solution: instead of |
1055 |
|
|
registering callbacks for each message type, C<Coro::MP> attached a |
1056 |
|
|
(coro-) thread to a port. The thread can then opt to selectively receive |
1057 |
|
|
messages it is interested in. Other messages are not lost, but queued, and |
1058 |
|
|
can be received at a later time. |
1059 |
|
|
|
1060 |
root |
1.43 |
The C<Coro::MP> module is not part of L<AnyEvent::MP>, but a separate |
1061 |
root |
1.42 |
module. It is, however, tightly integrated into C<AnyEvent::MP> - the |
1062 |
|
|
ports it creates are fully compatible to C<AnyEvent::MP> ports. |
1063 |
|
|
|
1064 |
|
|
In fact, C<Coro::MP> is more of an extension than a separate module: all |
1065 |
|
|
functions exported by C<AnyEvent::MP> are exported by it as well. |
1066 |
|
|
|
1067 |
|
|
To illustrate how programing with C<Coro::MP> looks like, consider the |
1068 |
|
|
following (slightly contrived) example: Let's implement a server that |
1069 |
|
|
accepts a C<< (write_file =>, $port, $path) >> message with a (source) |
1070 |
|
|
port and a filename, followed by as many C<< (data => $port, $data) >> |
1071 |
|
|
messages as required to fill the file, followed by an empty C<< (data => |
1072 |
|
|
$port) >> message. |
1073 |
|
|
|
1074 |
|
|
The server only writes a single file at a time, other requests will stay |
1075 |
|
|
in the queue until the current file has been finished. |
1076 |
|
|
|
1077 |
|
|
Here is an example implementation that uses L<Coro::AIO> and largely |
1078 |
|
|
ignores error handling: |
1079 |
|
|
|
1080 |
|
|
my $ioserver = port_async { |
1081 |
|
|
while () { |
1082 |
|
|
my ($tag, $port, $path) = get_cond; |
1083 |
|
|
|
1084 |
|
|
$tag eq "write_file" |
1085 |
|
|
or die "only write_file messages expected"; |
1086 |
|
|
|
1087 |
|
|
my $fh = aio_open $path, O_WRONLY|O_CREAT, 0666 |
1088 |
|
|
or die "$path: $!"; |
1089 |
|
|
|
1090 |
|
|
while () { |
1091 |
|
|
my (undef, undef, $data) = get_cond { |
1092 |
|
|
$_[0] eq "data" && $_[1] eq $port |
1093 |
|
|
} 5 |
1094 |
|
|
or die "timeout waiting for data message from $port\n"; |
1095 |
|
|
|
1096 |
|
|
length $data or last; |
1097 |
|
|
|
1098 |
|
|
aio_write $fh, undef, undef, $data, 0; |
1099 |
|
|
}; |
1100 |
|
|
} |
1101 |
|
|
}; |
1102 |
|
|
|
1103 |
|
|
mon $ioserver, sub { |
1104 |
|
|
warn "ioserver was killed: @_\n"; |
1105 |
|
|
}; |
1106 |
|
|
|
1107 |
|
|
Let's go through it part by part. |
1108 |
|
|
|
1109 |
|
|
my $ioserver = port_async { |
1110 |
|
|
|
1111 |
root |
1.43 |
Ports can be created by attaching a thread to an existing port via |
1112 |
|
|
C<rcv_async>, or as here by calling C<port_async> with the code to execute |
1113 |
root |
1.42 |
as a thread. The C<async> component comes from the fact that threads are |
1114 |
|
|
created using the C<Coro::async> function. |
1115 |
|
|
|
1116 |
|
|
The thread runs in a normal port context (so C<$SELF> is set). In |
1117 |
|
|
addition, when the thread returns, it will be C<kil> I<normally>, i.e. |
1118 |
|
|
without a reason argument. |
1119 |
|
|
|
1120 |
|
|
while () { |
1121 |
|
|
my ($tag, $port, $path) = get_cond; |
1122 |
|
|
or die "only write_file messages expected"; |
1123 |
|
|
|
1124 |
|
|
The thread is supposed to serve many file writes, which is why it executes |
1125 |
|
|
in a loop. The first thing it does is fetch the next message, using |
1126 |
|
|
C<get_cond>, the "conditional message get". Without a condition, it simply |
1127 |
|
|
fetches the next message from the queue, which I<must> be a C<write_file> |
1128 |
|
|
message. |
1129 |
|
|
|
1130 |
|
|
The message contains the C<$path> to the file, which is then created: |
1131 |
|
|
|
1132 |
|
|
my $fh = aio_open $path, O_WRONLY|O_CREAT, 0666 |
1133 |
|
|
or die "$path: $!"; |
1134 |
|
|
|
1135 |
|
|
Then we enter a loop again, to serve as many C<data> messages as |
1136 |
root |
1.43 |
necessary: |
1137 |
root |
1.42 |
|
1138 |
|
|
while () { |
1139 |
|
|
my (undef, undef, $data) = get_cond { |
1140 |
|
|
$_[0] eq "data" && $_[1] eq $port |
1141 |
|
|
} 5 |
1142 |
|
|
or die "timeout waiting for data message from $port\n"; |
1143 |
|
|
|
1144 |
|
|
This time, the condition is not empty, but instead a code block: similarly |
1145 |
|
|
to grep, the code block will be called with C<@_> set to each message in |
1146 |
|
|
the queue, and it has to return whether it wants to receive the message or |
1147 |
|
|
not. |
1148 |
|
|
|
1149 |
|
|
In this case we are interested in C<data> messages (C<< $_[0] eq "data" |
1150 |
|
|
>>), whose first element is the source port (C<< $_[1] eq $port >>). |
1151 |
|
|
|
1152 |
|
|
The condition must be this strict, as it is possible to receive both |
1153 |
|
|
C<write_file> messages and C<data> messages from other ports while we |
1154 |
|
|
handle the file writing. |
1155 |
|
|
|
1156 |
|
|
The lone C<5> at the end is a timeout - when no matching message is |
1157 |
|
|
received within C<5> seconds, we assume an error and C<die>. |
1158 |
|
|
|
1159 |
|
|
When an empty C<data> message is received we are done and can close the |
1160 |
|
|
file (which is done automatically as C<$fh> goes out of scope): |
1161 |
|
|
|
1162 |
|
|
length $data or last; |
1163 |
|
|
|
1164 |
|
|
Otherwise we need to write the data: |
1165 |
|
|
|
1166 |
|
|
aio_write $fh, undef, undef, $data, 0; |
1167 |
|
|
|
1168 |
root |
1.43 |
That's basically it. Note that every process should have some kind of |
1169 |
root |
1.42 |
supervisor. In our case, the supervisor simply prints any error message: |
1170 |
|
|
|
1171 |
|
|
mon $ioserver, sub { |
1172 |
|
|
warn "ioserver was killed: @_\n"; |
1173 |
|
|
}; |
1174 |
|
|
|
1175 |
|
|
Here is a usage example: |
1176 |
|
|
|
1177 |
|
|
port_async { |
1178 |
|
|
snd $ioserver, write_file => $SELF, "/tmp/unsafe"; |
1179 |
|
|
snd $ioserver, data => $SELF, "abc\n"; |
1180 |
|
|
snd $ioserver, data => $SELF, "def\n"; |
1181 |
|
|
snd $ioserver, data => $SELF; |
1182 |
|
|
}; |
1183 |
|
|
|
1184 |
|
|
The messages are sent without any flow control or acknowledgement (feel |
1185 |
|
|
free to improve). Also, the source port does not actually need to be a |
1186 |
|
|
port - any unique ID will do - but port identifiers happen to be a simple |
1187 |
|
|
source of network-wide unique IDs. |
1188 |
|
|
|
1189 |
|
|
Apart from C<get_cond> as seen above, there are other ways to receive |
1190 |
|
|
messages. The C<write_file> message above could also selectively be |
1191 |
|
|
received using a C<get> call: |
1192 |
|
|
|
1193 |
|
|
my ($port, $path) = get "write_file"; |
1194 |
|
|
|
1195 |
|
|
This is simpler, but when some other code part sends an unexpected message |
1196 |
|
|
to the C<$ioserver> it will stay in the queue forever. As a rule of thumb, |
1197 |
|
|
every threaded port should have a "fetch next message unconditionally" |
1198 |
|
|
somewhere, to avoid filling up the queue. |
1199 |
|
|
|
1200 |
|
|
It is also possible to switch-like C<get_conds>: |
1201 |
|
|
|
1202 |
|
|
get_cond { |
1203 |
|
|
$_[0] eq "msg1" and return sub { |
1204 |
|
|
my (undef, @msg1_data) = @_; |
1205 |
|
|
...; |
1206 |
|
|
}; |
1207 |
|
|
|
1208 |
|
|
$_[0] eq "msg2" and return sub { |
1209 |
|
|
my (undef, @msg2_data) = @_; |
1210 |
|
|
...; |
1211 |
|
|
}; |
1212 |
|
|
|
1213 |
|
|
die "unexpected message $_[0] received"; |
1214 |
|
|
}; |
1215 |
|
|
|
1216 |
root |
1.37 |
=head1 THE END |
1217 |
|
|
|
1218 |
|
|
This is the end of this introduction, but hopefully not the end of |
1219 |
root |
1.43 |
your career as AEMP user. I hope the tutorial was enough to make the |
1220 |
root |
1.37 |
basic concepts clear. Keep in mind that distributed programming is not |
1221 |
|
|
completely trivial, that AnyEvent::MP is still in it's infancy, and I hope |
1222 |
|
|
it will be useful to create exciting new applications. |
1223 |
|
|
|
1224 |
elmex |
1.1 |
=head1 SEE ALSO |
1225 |
|
|
|
1226 |
|
|
L<AnyEvent::MP> |
1227 |
|
|
|
1228 |
elmex |
1.20 |
L<AnyEvent::MP::Global> |
1229 |
|
|
|
1230 |
root |
1.42 |
L<Coro::MP> |
1231 |
|
|
|
1232 |
root |
1.34 |
L<AnyEvent> |
1233 |
|
|
|
1234 |
elmex |
1.1 |
=head1 AUTHOR |
1235 |
|
|
|
1236 |
|
|
Robin Redeker <elmex@ta-sa.org> |
1237 |
root |
1.32 |
Marc Lehmann <schmorp@schmorp.de> |
1238 |
root |
1.4 |
|