1 |
=head1 NAME |
2 |
|
3 |
AnyEvent::MP - multi-processing/message-passing framework |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use AnyEvent::MP; |
8 |
|
9 |
$NODE # contains this node's node ID |
10 |
NODE # returns this node's node ID |
11 |
|
12 |
$SELF # receiving/own port id in rcv callbacks |
13 |
|
14 |
# initialise the node so it can send/receive messages |
15 |
configure; |
16 |
|
17 |
# ports are message destinations |
18 |
|
19 |
# sending messages |
20 |
snd $port, type => data...; |
21 |
snd $port, @msg; |
22 |
snd @msg_with_first_element_being_a_port; |
23 |
|
24 |
# creating/using ports, the simple way |
25 |
my $simple_port = port { my @msg = @_ }; |
26 |
|
27 |
# creating/using ports, tagged message matching |
28 |
my $port = port; |
29 |
rcv $port, ping => sub { snd $_[0], "pong" }; |
30 |
rcv $port, pong => sub { warn "pong received\n" }; |
31 |
|
32 |
# create a port on another node |
33 |
my $port = spawn $node, $initfunc, @initdata; |
34 |
|
35 |
# monitoring |
36 |
mon $port, $cb->(@msg) # callback is invoked on death |
37 |
mon $port, $otherport # kill otherport on abnormal death |
38 |
mon $port, $otherport, @msg # send message on death |
39 |
|
40 |
=head1 CURRENT STATUS |
41 |
|
42 |
bin/aemp - stable. |
43 |
AnyEvent::MP - stable API, should work. |
44 |
AnyEvent::MP::Intro - explains most concepts. |
45 |
AnyEvent::MP::Kernel - mostly stable. |
46 |
AnyEvent::MP::Global - stable but incomplete, protocol not yet final. |
47 |
|
48 |
stay tuned. |
49 |
|
50 |
=head1 DESCRIPTION |
51 |
|
52 |
This module (-family) implements a simple message passing framework. |
53 |
|
54 |
Despite its simplicity, you can securely message other processes running |
55 |
on the same or other hosts, and you can supervise entities remotely. |
56 |
|
57 |
For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
58 |
manual page and the examples under F<eg/>. |
59 |
|
60 |
=head1 CONCEPTS |
61 |
|
62 |
=over 4 |
63 |
|
64 |
=item port |
65 |
|
66 |
Not to be confused with a TCP port, a "port" is something you can send |
67 |
messages to (with the C<snd> function). |
68 |
|
69 |
Ports allow you to register C<rcv> handlers that can match all or just |
70 |
some messages. Messages send to ports will not be queued, regardless of |
71 |
anything was listening for them or not. |
72 |
|
73 |
=item port ID - C<nodeid#portname> |
74 |
|
75 |
A port ID is the concatenation of a node ID, a hash-mark (C<#>) as |
76 |
separator, and a port name (a printable string of unspecified format). |
77 |
|
78 |
=item node |
79 |
|
80 |
A node is a single process containing at least one port - the node port, |
81 |
which enables nodes to manage each other remotely, and to create new |
82 |
ports. |
83 |
|
84 |
Nodes are either public (have one or more listening ports) or private |
85 |
(no listening ports). Private nodes cannot talk to other private nodes |
86 |
currently. |
87 |
|
88 |
=item node ID - C<[a-za-Z0-9_\-.:]+> |
89 |
|
90 |
A node ID is a string that uniquely identifies the node within a |
91 |
network. Depending on the configuration used, node IDs can look like a |
92 |
hostname, a hostname and a port, or a random string. AnyEvent::MP itself |
93 |
doesn't interpret node IDs in any way. |
94 |
|
95 |
=item binds - C<ip:port> |
96 |
|
97 |
Nodes can only talk to each other by creating some kind of connection to |
98 |
each other. To do this, nodes should listen on one or more local transport |
99 |
endpoints - binds. Currently, only standard C<ip:port> specifications can |
100 |
be used, which specify TCP ports to listen on. |
101 |
|
102 |
=item seed nodes |
103 |
|
104 |
When a node starts, it knows nothing about the network. To teach the node |
105 |
about the network it first has to contact some other node within the |
106 |
network. This node is called a seed. |
107 |
|
108 |
Apart from the fact that other nodes know them as seed nodes and they have |
109 |
to have fixed listening addresses, seed nodes are perfectly normal nodes - |
110 |
any node can function as a seed node for others. |
111 |
|
112 |
In addition to discovering the network, seed nodes are also used to |
113 |
maintain the network and to connect nodes that otherwise would have |
114 |
trouble connecting. They form the backbone of the AnyEvent::MP network. |
115 |
|
116 |
Seed nodes are expected to be long-running, and at least one seed node |
117 |
should always be available. |
118 |
|
119 |
=item seeds - C<host:port> |
120 |
|
121 |
Seeds are transport endpoint(s) (usually a hostname/IP address and a |
122 |
TCP port) of nodes thta should be used as seed nodes. |
123 |
|
124 |
The nodes listening on those endpoints are expected to be long-running, |
125 |
and at least one of those should always be available. When nodes run out |
126 |
of connections (e.g. due to a network error), they try to re-establish |
127 |
connections to some seednodes again to join the network. |
128 |
|
129 |
=back |
130 |
|
131 |
=head1 VARIABLES/FUNCTIONS |
132 |
|
133 |
=over 4 |
134 |
|
135 |
=cut |
136 |
|
137 |
package AnyEvent::MP; |
138 |
|
139 |
use AnyEvent::MP::Kernel; |
140 |
|
141 |
use common::sense; |
142 |
|
143 |
use Carp (); |
144 |
|
145 |
use AE (); |
146 |
|
147 |
use base "Exporter"; |
148 |
|
149 |
our $VERSION = $AnyEvent::MP::Kernel::VERSION; |
150 |
|
151 |
our @EXPORT = qw( |
152 |
NODE $NODE *SELF node_of after |
153 |
configure |
154 |
snd rcv mon mon_guard kil reg psub spawn |
155 |
port |
156 |
); |
157 |
|
158 |
our $SELF; |
159 |
|
160 |
sub _self_die() { |
161 |
my $msg = $@; |
162 |
$msg =~ s/\n+$// unless ref $msg; |
163 |
kil $SELF, die => $msg; |
164 |
} |
165 |
|
166 |
=item $thisnode = NODE / $NODE |
167 |
|
168 |
The C<NODE> function returns, and the C<$NODE> variable contains, the node |
169 |
ID of the node running in the current process. This value is initialised by |
170 |
a call to C<configure>. |
171 |
|
172 |
=item $nodeid = node_of $port |
173 |
|
174 |
Extracts and returns the node ID from a port ID or a node ID. |
175 |
|
176 |
=item configure $profile, key => value... |
177 |
|
178 |
=item configure key => value... |
179 |
|
180 |
Before a node can talk to other nodes on the network (i.e. enter |
181 |
"distributed mode") it has to configure itself - the minimum a node needs |
182 |
to know is its own name, and optionally it should know the addresses of |
183 |
some other nodes in the network to discover other nodes. |
184 |
|
185 |
This function configures a node - it must be called exactly once (or |
186 |
never) before calling other AnyEvent::MP functions. |
187 |
|
188 |
=over 4 |
189 |
|
190 |
=item step 1, gathering configuration from profiles |
191 |
|
192 |
The function first looks up a profile in the aemp configuration (see the |
193 |
L<aemp> commandline utility). The profile name can be specified via the |
194 |
named C<profile> parameter or can simply be the first parameter). If it is |
195 |
missing, then the nodename (F<uname -n>) will be used as profile name. |
196 |
|
197 |
The profile data is then gathered as follows: |
198 |
|
199 |
First, all remaining key => value pairs (all of which are conveniently |
200 |
undocumented at the moment) will be interpreted as configuration |
201 |
data. Then they will be overwritten by any values specified in the global |
202 |
default configuration (see the F<aemp> utility), then the chain of |
203 |
profiles chosen by the profile name (and any C<parent> attributes). |
204 |
|
205 |
That means that the values specified in the profile have highest priority |
206 |
and the values specified directly via C<configure> have lowest priority, |
207 |
and can only be used to specify defaults. |
208 |
|
209 |
If the profile specifies a node ID, then this will become the node ID of |
210 |
this process. If not, then the profile name will be used as node ID. The |
211 |
special node ID of C<anon/> will be replaced by a random node ID. |
212 |
|
213 |
=item step 2, bind listener sockets |
214 |
|
215 |
The next step is to look up the binds in the profile, followed by binding |
216 |
aemp protocol listeners on all binds specified (it is possible and valid |
217 |
to have no binds, meaning that the node cannot be contacted form the |
218 |
outside. This means the node cannot talk to other nodes that also have no |
219 |
binds, but it can still talk to all "normal" nodes). |
220 |
|
221 |
If the profile does not specify a binds list, then a default of C<*> is |
222 |
used, meaning the node will bind on a dynamically-assigned port on every |
223 |
local IP address it finds. |
224 |
|
225 |
=item step 3, connect to seed nodes |
226 |
|
227 |
As the last step, the seeds list from the profile is passed to the |
228 |
L<AnyEvent::MP::Global> module, which will then use it to keep |
229 |
connectivity with at least one node at any point in time. |
230 |
|
231 |
=back |
232 |
|
233 |
Example: become a distributed node using the locla node name as profile. |
234 |
This should be the most common form of invocation for "daemon"-type nodes. |
235 |
|
236 |
configure |
237 |
|
238 |
Example: become an anonymous node. This form is often used for commandline |
239 |
clients. |
240 |
|
241 |
configure nodeid => "anon/"; |
242 |
|
243 |
Example: configure a node using a profile called seed, which si suitable |
244 |
for a seed node as it binds on all local addresses on a fixed port (4040, |
245 |
customary for aemp). |
246 |
|
247 |
# use the aemp commandline utility |
248 |
# aemp profile seed nodeid anon/ binds '*:4040' |
249 |
|
250 |
# then use it |
251 |
configure profile => "seed"; |
252 |
|
253 |
# or simply use aemp from the shell again: |
254 |
# aemp run profile seed |
255 |
|
256 |
# or provide a nicer-to-remember nodeid |
257 |
# aemp run profile seed nodeid "$(hostname)" |
258 |
|
259 |
=item $SELF |
260 |
|
261 |
Contains the current port id while executing C<rcv> callbacks or C<psub> |
262 |
blocks. |
263 |
|
264 |
=item *SELF, SELF, %SELF, @SELF... |
265 |
|
266 |
Due to some quirks in how perl exports variables, it is impossible to |
267 |
just export C<$SELF>, all the symbols named C<SELF> are exported by this |
268 |
module, but only C<$SELF> is currently used. |
269 |
|
270 |
=item snd $port, type => @data |
271 |
|
272 |
=item snd $port, @msg |
273 |
|
274 |
Send the given message to the given port, which can identify either a |
275 |
local or a remote port, and must be a port ID. |
276 |
|
277 |
While the message can be almost anything, it is highly recommended to |
278 |
use a string as first element (a port ID, or some word that indicates a |
279 |
request type etc.) and to consist if only simple perl values (scalars, |
280 |
arrays, hashes) - if you think you need to pass an object, think again. |
281 |
|
282 |
The message data logically becomes read-only after a call to this |
283 |
function: modifying any argument (or values referenced by them) is |
284 |
forbidden, as there can be considerable time between the call to C<snd> |
285 |
and the time the message is actually being serialised - in fact, it might |
286 |
never be copied as within the same process it is simply handed to the |
287 |
receiving port. |
288 |
|
289 |
The type of data you can transfer depends on the transport protocol: when |
290 |
JSON is used, then only strings, numbers and arrays and hashes consisting |
291 |
of those are allowed (no objects). When Storable is used, then anything |
292 |
that Storable can serialise and deserialise is allowed, and for the local |
293 |
node, anything can be passed. Best rely only on the common denominator of |
294 |
these. |
295 |
|
296 |
=item $local_port = port |
297 |
|
298 |
Create a new local port object and returns its port ID. Initially it has |
299 |
no callbacks set and will throw an error when it receives messages. |
300 |
|
301 |
=item $local_port = port { my @msg = @_ } |
302 |
|
303 |
Creates a new local port, and returns its ID. Semantically the same as |
304 |
creating a port and calling C<rcv $port, $callback> on it. |
305 |
|
306 |
The block will be called for every message received on the port, with the |
307 |
global variable C<$SELF> set to the port ID. Runtime errors will cause the |
308 |
port to be C<kil>ed. The message will be passed as-is, no extra argument |
309 |
(i.e. no port ID) will be passed to the callback. |
310 |
|
311 |
If you want to stop/destroy the port, simply C<kil> it: |
312 |
|
313 |
my $port = port { |
314 |
my @msg = @_; |
315 |
... |
316 |
kil $SELF; |
317 |
}; |
318 |
|
319 |
=cut |
320 |
|
321 |
sub rcv($@); |
322 |
|
323 |
sub _kilme { |
324 |
die "received message on port without callback"; |
325 |
} |
326 |
|
327 |
sub port(;&) { |
328 |
my $id = "$UNIQ." . $ID++; |
329 |
my $port = "$NODE#$id"; |
330 |
|
331 |
rcv $port, shift || \&_kilme; |
332 |
|
333 |
$port |
334 |
} |
335 |
|
336 |
=item rcv $local_port, $callback->(@msg) |
337 |
|
338 |
Replaces the default callback on the specified port. There is no way to |
339 |
remove the default callback: use C<sub { }> to disable it, or better |
340 |
C<kil> the port when it is no longer needed. |
341 |
|
342 |
The global C<$SELF> (exported by this module) contains C<$port> while |
343 |
executing the callback. Runtime errors during callback execution will |
344 |
result in the port being C<kil>ed. |
345 |
|
346 |
The default callback received all messages not matched by a more specific |
347 |
C<tag> match. |
348 |
|
349 |
=item rcv $local_port, tag => $callback->(@msg_without_tag), ... |
350 |
|
351 |
Register (or replace) callbacks to be called on messages starting with the |
352 |
given tag on the given port (and return the port), or unregister it (when |
353 |
C<$callback> is C<$undef> or missing). There can only be one callback |
354 |
registered for each tag. |
355 |
|
356 |
The original message will be passed to the callback, after the first |
357 |
element (the tag) has been removed. The callback will use the same |
358 |
environment as the default callback (see above). |
359 |
|
360 |
Example: create a port and bind receivers on it in one go. |
361 |
|
362 |
my $port = rcv port, |
363 |
msg1 => sub { ... }, |
364 |
msg2 => sub { ... }, |
365 |
; |
366 |
|
367 |
Example: create a port, bind receivers and send it in a message elsewhere |
368 |
in one go: |
369 |
|
370 |
snd $otherport, reply => |
371 |
rcv port, |
372 |
msg1 => sub { ... }, |
373 |
... |
374 |
; |
375 |
|
376 |
Example: temporarily register a rcv callback for a tag matching some port |
377 |
(e.g. for a rpc reply) and unregister it after a message was received. |
378 |
|
379 |
rcv $port, $otherport => sub { |
380 |
my @reply = @_; |
381 |
|
382 |
rcv $SELF, $otherport; |
383 |
}; |
384 |
|
385 |
=cut |
386 |
|
387 |
sub rcv($@) { |
388 |
my $port = shift; |
389 |
my ($nodeid, $portid) = split /#/, $port, 2; |
390 |
|
391 |
$NODE{$nodeid} == $NODE{""} |
392 |
or Carp::croak "$port: rcv can only be called on local ports, caught"; |
393 |
|
394 |
while (@_) { |
395 |
if (ref $_[0]) { |
396 |
if (my $self = $PORT_DATA{$portid}) { |
397 |
"AnyEvent::MP::Port" eq ref $self |
398 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
399 |
|
400 |
$self->[2] = shift; |
401 |
} else { |
402 |
my $cb = shift; |
403 |
$PORT{$portid} = sub { |
404 |
local $SELF = $port; |
405 |
eval { &$cb }; _self_die if $@; |
406 |
}; |
407 |
} |
408 |
} elsif (defined $_[0]) { |
409 |
my $self = $PORT_DATA{$portid} ||= do { |
410 |
my $self = bless [$PORT{$port} || sub { }, { }, $port], "AnyEvent::MP::Port"; |
411 |
|
412 |
$PORT{$portid} = sub { |
413 |
local $SELF = $port; |
414 |
|
415 |
if (my $cb = $self->[1]{$_[0]}) { |
416 |
shift; |
417 |
eval { &$cb }; _self_die if $@; |
418 |
} else { |
419 |
&{ $self->[0] }; |
420 |
} |
421 |
}; |
422 |
|
423 |
$self |
424 |
}; |
425 |
|
426 |
"AnyEvent::MP::Port" eq ref $self |
427 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
428 |
|
429 |
my ($tag, $cb) = splice @_, 0, 2; |
430 |
|
431 |
if (defined $cb) { |
432 |
$self->[1]{$tag} = $cb; |
433 |
} else { |
434 |
delete $self->[1]{$tag}; |
435 |
} |
436 |
} |
437 |
} |
438 |
|
439 |
$port |
440 |
} |
441 |
|
442 |
=item $closure = psub { BLOCK } |
443 |
|
444 |
Remembers C<$SELF> and creates a closure out of the BLOCK. When the |
445 |
closure is executed, sets up the environment in the same way as in C<rcv> |
446 |
callbacks, i.e. runtime errors will cause the port to get C<kil>ed. |
447 |
|
448 |
This is useful when you register callbacks from C<rcv> callbacks: |
449 |
|
450 |
rcv delayed_reply => sub { |
451 |
my ($delay, @reply) = @_; |
452 |
my $timer = AE::timer $delay, 0, psub { |
453 |
snd @reply, $SELF; |
454 |
}; |
455 |
}; |
456 |
|
457 |
=cut |
458 |
|
459 |
sub psub(&) { |
460 |
my $cb = shift; |
461 |
|
462 |
my $port = $SELF |
463 |
or Carp::croak "psub can only be called from within rcv or psub callbacks, not"; |
464 |
|
465 |
sub { |
466 |
local $SELF = $port; |
467 |
|
468 |
if (wantarray) { |
469 |
my @res = eval { &$cb }; |
470 |
_self_die if $@; |
471 |
@res |
472 |
} else { |
473 |
my $res = eval { &$cb }; |
474 |
_self_die if $@; |
475 |
$res |
476 |
} |
477 |
} |
478 |
} |
479 |
|
480 |
=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies |
481 |
|
482 |
=item $guard = mon $port, $rcvport # kill $rcvport when $port dies |
483 |
|
484 |
=item $guard = mon $port # kill $SELF when $port dies |
485 |
|
486 |
=item $guard = mon $port, $rcvport, @msg # send a message when $port dies |
487 |
|
488 |
Monitor the given port and do something when the port is killed or |
489 |
messages to it were lost, and optionally return a guard that can be used |
490 |
to stop monitoring again. |
491 |
|
492 |
In the first form (callback), the callback is simply called with any |
493 |
number of C<@reason> elements (no @reason means that the port was deleted |
494 |
"normally"). Note also that I<< the callback B<must> never die >>, so use |
495 |
C<eval> if unsure. |
496 |
|
497 |
In the second form (another port given), the other port (C<$rcvport>) |
498 |
will be C<kil>'ed with C<@reason>, if a @reason was specified, i.e. on |
499 |
"normal" kils nothing happens, while under all other conditions, the other |
500 |
port is killed with the same reason. |
501 |
|
502 |
The third form (kill self) is the same as the second form, except that |
503 |
C<$rvport> defaults to C<$SELF>. |
504 |
|
505 |
In the last form (message), a message of the form C<@msg, @reason> will be |
506 |
C<snd>. |
507 |
|
508 |
Monitoring-actions are one-shot: once messages are lost (and a monitoring |
509 |
alert was raised), they are removed and will not trigger again. |
510 |
|
511 |
As a rule of thumb, monitoring requests should always monitor a port from |
512 |
a local port (or callback). The reason is that kill messages might get |
513 |
lost, just like any other message. Another less obvious reason is that |
514 |
even monitoring requests can get lost (for example, when the connection |
515 |
to the other node goes down permanently). When monitoring a port locally |
516 |
these problems do not exist. |
517 |
|
518 |
C<mon> effectively guarantees that, in the absence of hardware failures, |
519 |
after starting the monitor, either all messages sent to the port will |
520 |
arrive, or the monitoring action will be invoked after possible message |
521 |
loss has been detected. No messages will be lost "in between" (after |
522 |
the first lost message no further messages will be received by the |
523 |
port). After the monitoring action was invoked, further messages might get |
524 |
delivered again. |
525 |
|
526 |
Inter-host-connection timeouts and monitoring depend on the transport |
527 |
used. The only transport currently implemented is TCP, and AnyEvent::MP |
528 |
relies on TCP to detect node-downs (this can take 10-15 minutes on a |
529 |
non-idle connection, and usually around two hours for idle conenctions). |
530 |
|
531 |
This means that monitoring is good for program errors and cleaning up |
532 |
stuff eventually, but they are no replacement for a timeout when you need |
533 |
to ensure some maximum latency. |
534 |
|
535 |
Example: call a given callback when C<$port> is killed. |
536 |
|
537 |
mon $port, sub { warn "port died because of <@_>\n" }; |
538 |
|
539 |
Example: kill ourselves when C<$port> is killed abnormally. |
540 |
|
541 |
mon $port; |
542 |
|
543 |
Example: send us a restart message when another C<$port> is killed. |
544 |
|
545 |
mon $port, $self => "restart"; |
546 |
|
547 |
=cut |
548 |
|
549 |
sub mon { |
550 |
my ($nodeid, $port) = split /#/, shift, 2; |
551 |
|
552 |
my $node = $NODE{$nodeid} || add_node $nodeid; |
553 |
|
554 |
my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,'; |
555 |
|
556 |
unless (ref $cb) { |
557 |
if (@_) { |
558 |
# send a kill info message |
559 |
my (@msg) = ($cb, @_); |
560 |
$cb = sub { snd @msg, @_ }; |
561 |
} else { |
562 |
# simply kill other port |
563 |
my $port = $cb; |
564 |
$cb = sub { kil $port, @_ if @_ }; |
565 |
} |
566 |
} |
567 |
|
568 |
$node->monitor ($port, $cb); |
569 |
|
570 |
defined wantarray |
571 |
and AnyEvent::Util::guard { $node->unmonitor ($port, $cb) } |
572 |
} |
573 |
|
574 |
=item $guard = mon_guard $port, $ref, $ref... |
575 |
|
576 |
Monitors the given C<$port> and keeps the passed references. When the port |
577 |
is killed, the references will be freed. |
578 |
|
579 |
Optionally returns a guard that will stop the monitoring. |
580 |
|
581 |
This function is useful when you create e.g. timers or other watchers and |
582 |
want to free them when the port gets killed (note the use of C<psub>): |
583 |
|
584 |
$port->rcv (start => sub { |
585 |
my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub { |
586 |
undef $timer if 0.9 < rand; |
587 |
}); |
588 |
}); |
589 |
|
590 |
=cut |
591 |
|
592 |
sub mon_guard { |
593 |
my ($port, @refs) = @_; |
594 |
|
595 |
#TODO: mon-less form? |
596 |
|
597 |
mon $port, sub { 0 && @refs } |
598 |
} |
599 |
|
600 |
=item kil $port[, @reason] |
601 |
|
602 |
Kill the specified port with the given C<@reason>. |
603 |
|
604 |
If no C<@reason> is specified, then the port is killed "normally" (ports |
605 |
monitoring other ports will not necessarily die because a port dies |
606 |
"normally"). |
607 |
|
608 |
Otherwise, linked ports get killed with the same reason (second form of |
609 |
C<mon>, see above). |
610 |
|
611 |
Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks |
612 |
will be reported as reason C<< die => $@ >>. |
613 |
|
614 |
Transport/communication errors are reported as C<< transport_error => |
615 |
$message >>. |
616 |
|
617 |
=cut |
618 |
|
619 |
=item $port = spawn $node, $initfunc[, @initdata] |
620 |
|
621 |
Creates a port on the node C<$node> (which can also be a port ID, in which |
622 |
case it's the node where that port resides). |
623 |
|
624 |
The port ID of the newly created port is returned immediately, and it is |
625 |
possible to immediately start sending messages or to monitor the port. |
626 |
|
627 |
After the port has been created, the init function is called on the remote |
628 |
node, in the same context as a C<rcv> callback. This function must be a |
629 |
fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To |
630 |
specify a function in the main program, use C<::name>. |
631 |
|
632 |
If the function doesn't exist, then the node tries to C<require> |
633 |
the package, then the package above the package and so on (e.g. |
634 |
C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
635 |
exists or it runs out of package names. |
636 |
|
637 |
The init function is then called with the newly-created port as context |
638 |
object (C<$SELF>) and the C<@initdata> values as arguments. It I<must> |
639 |
call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise |
640 |
the port might not get created. |
641 |
|
642 |
A common idiom is to pass a local port, immediately monitor the spawned |
643 |
port, and in the remote init function, immediately monitor the passed |
644 |
local port. This two-way monitoring ensures that both ports get cleaned up |
645 |
when there is a problem. |
646 |
|
647 |
C<spawn> guarantees that the C<$initfunc> has no visible effects on the |
648 |
caller before C<spawn> returns (by delaying invocation when spawn is |
649 |
called for the local node). |
650 |
|
651 |
Example: spawn a chat server port on C<$othernode>. |
652 |
|
653 |
# this node, executed from within a port context: |
654 |
my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
655 |
mon $server; |
656 |
|
657 |
# init function on C<$othernode> |
658 |
sub connect { |
659 |
my ($srcport) = @_; |
660 |
|
661 |
mon $srcport; |
662 |
|
663 |
rcv $SELF, sub { |
664 |
... |
665 |
}; |
666 |
} |
667 |
|
668 |
=cut |
669 |
|
670 |
sub _spawn { |
671 |
my $port = shift; |
672 |
my $init = shift; |
673 |
|
674 |
# rcv will create the actual port |
675 |
local $SELF = "$NODE#$port"; |
676 |
eval { |
677 |
&{ load_func $init } |
678 |
}; |
679 |
_self_die if $@; |
680 |
} |
681 |
|
682 |
sub spawn(@) { |
683 |
my ($nodeid, undef) = split /#/, shift, 2; |
684 |
|
685 |
my $id = "$RUNIQ." . $ID++; |
686 |
|
687 |
$_[0] =~ /::/ |
688 |
or Carp::croak "spawn init function must be a fully-qualified name, caught"; |
689 |
|
690 |
snd_to_func $nodeid, "AnyEvent::MP::_spawn" => $id, @_; |
691 |
|
692 |
"$nodeid#$id" |
693 |
} |
694 |
|
695 |
=item after $timeout, @msg |
696 |
|
697 |
=item after $timeout, $callback |
698 |
|
699 |
Either sends the given message, or call the given callback, after the |
700 |
specified number of seconds. |
701 |
|
702 |
This is simply a utility function that comes in handy at times - the |
703 |
AnyEvent::MP author is not convinced of the wisdom of having it, though, |
704 |
so it may go away in the future. |
705 |
|
706 |
=cut |
707 |
|
708 |
sub after($@) { |
709 |
my ($timeout, @action) = @_; |
710 |
|
711 |
my $t; $t = AE::timer $timeout, 0, sub { |
712 |
undef $t; |
713 |
ref $action[0] |
714 |
? $action[0]() |
715 |
: snd @action; |
716 |
}; |
717 |
} |
718 |
|
719 |
=back |
720 |
|
721 |
=head1 AnyEvent::MP vs. Distributed Erlang |
722 |
|
723 |
AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
724 |
== aemp node, Erlang process == aemp port), so many of the documents and |
725 |
programming techniques employed by Erlang apply to AnyEvent::MP. Here is a |
726 |
sample: |
727 |
|
728 |
http://www.Erlang.se/doc/programming_rules.shtml |
729 |
http://Erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4 |
730 |
http://Erlang.org/download/Erlang-book-part1.pdf # chapters 5 and 6 |
731 |
http://Erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5 |
732 |
|
733 |
Despite the similarities, there are also some important differences: |
734 |
|
735 |
=over 4 |
736 |
|
737 |
=item * Node IDs are arbitrary strings in AEMP. |
738 |
|
739 |
Erlang relies on special naming and DNS to work everywhere in the same |
740 |
way. AEMP relies on each node somehow knowing its own address(es) (e.g. by |
741 |
configuration or DNS), but will otherwise discover other odes itself. |
742 |
|
743 |
=item * Erlang has a "remote ports are like local ports" philosophy, AEMP |
744 |
uses "local ports are like remote ports". |
745 |
|
746 |
The failure modes for local ports are quite different (runtime errors |
747 |
only) then for remote ports - when a local port dies, you I<know> it dies, |
748 |
when a connection to another node dies, you know nothing about the other |
749 |
port. |
750 |
|
751 |
Erlang pretends remote ports are as reliable as local ports, even when |
752 |
they are not. |
753 |
|
754 |
AEMP encourages a "treat remote ports differently" philosophy, with local |
755 |
ports being the special case/exception, where transport errors cannot |
756 |
occur. |
757 |
|
758 |
=item * Erlang uses processes and a mailbox, AEMP does not queue. |
759 |
|
760 |
Erlang uses processes that selectively receive messages, and therefore |
761 |
needs a queue. AEMP is event based, queuing messages would serve no |
762 |
useful purpose. For the same reason the pattern-matching abilities of |
763 |
AnyEvent::MP are more limited, as there is little need to be able to |
764 |
filter messages without dequeuing them. |
765 |
|
766 |
(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP). |
767 |
|
768 |
=item * Erlang sends are synchronous, AEMP sends are asynchronous. |
769 |
|
770 |
Sending messages in Erlang is synchronous and blocks the process (and |
771 |
so does not need a queue that can overflow). AEMP sends are immediate, |
772 |
connection establishment is handled in the background. |
773 |
|
774 |
=item * Erlang suffers from silent message loss, AEMP does not. |
775 |
|
776 |
Erlang makes few guarantees on messages delivery - messages can get lost |
777 |
without any of the processes realising it (i.e. you send messages a, b, |
778 |
and c, and the other side only receives messages a and c). |
779 |
|
780 |
AEMP guarantees correct ordering, and the guarantee that after one message |
781 |
is lost, all following ones sent to the same port are lost as well, until |
782 |
monitoring raises an error, so there are no silent "holes" in the message |
783 |
sequence. |
784 |
|
785 |
=item * Erlang can send messages to the wrong port, AEMP does not. |
786 |
|
787 |
In Erlang it is quite likely that a node that restarts reuses a process ID |
788 |
known to other nodes for a completely different process, causing messages |
789 |
destined for that process to end up in an unrelated process. |
790 |
|
791 |
AEMP never reuses port IDs, so old messages or old port IDs floating |
792 |
around in the network will not be sent to an unrelated port. |
793 |
|
794 |
=item * Erlang uses unprotected connections, AEMP uses secure |
795 |
authentication and can use TLS. |
796 |
|
797 |
AEMP can use a proven protocol - TLS - to protect connections and |
798 |
securely authenticate nodes. |
799 |
|
800 |
=item * The AEMP protocol is optimised for both text-based and binary |
801 |
communications. |
802 |
|
803 |
The AEMP protocol, unlike the Erlang protocol, supports both programming |
804 |
language independent text-only protocols (good for debugging) and binary, |
805 |
language-specific serialisers (e.g. Storable). By default, unless TLS is |
806 |
used, the protocol is actually completely text-based. |
807 |
|
808 |
It has also been carefully designed to be implementable in other languages |
809 |
with a minimum of work while gracefully degrading functionality to make the |
810 |
protocol simple. |
811 |
|
812 |
=item * AEMP has more flexible monitoring options than Erlang. |
813 |
|
814 |
In Erlang, you can chose to receive I<all> exit signals as messages |
815 |
or I<none>, there is no in-between, so monitoring single processes is |
816 |
difficult to implement. Monitoring in AEMP is more flexible than in |
817 |
Erlang, as one can choose between automatic kill, exit message or callback |
818 |
on a per-process basis. |
819 |
|
820 |
=item * Erlang tries to hide remote/local connections, AEMP does not. |
821 |
|
822 |
Monitoring in Erlang is not an indicator of process death/crashes, in the |
823 |
same way as linking is (except linking is unreliable in Erlang). |
824 |
|
825 |
In AEMP, you don't "look up" registered port names or send to named ports |
826 |
that might or might not be persistent. Instead, you normally spawn a port |
827 |
on the remote node. The init function monitors you, and you monitor the |
828 |
remote port. Since both monitors are local to the node, they are much more |
829 |
reliable (no need for C<spawn_link>). |
830 |
|
831 |
This also saves round-trips and avoids sending messages to the wrong port |
832 |
(hard to do in Erlang). |
833 |
|
834 |
=back |
835 |
|
836 |
=head1 RATIONALE |
837 |
|
838 |
=over 4 |
839 |
|
840 |
=item Why strings for port and node IDs, why not objects? |
841 |
|
842 |
We considered "objects", but found that the actual number of methods |
843 |
that can be called are quite low. Since port and node IDs travel over |
844 |
the network frequently, the serialising/deserialising would add lots of |
845 |
overhead, as well as having to keep a proxy object everywhere. |
846 |
|
847 |
Strings can easily be printed, easily serialised etc. and need no special |
848 |
procedures to be "valid". |
849 |
|
850 |
And as a result, a miniport consists of a single closure stored in a |
851 |
global hash - it can't become much cheaper. |
852 |
|
853 |
=item Why favour JSON, why not a real serialising format such as Storable? |
854 |
|
855 |
In fact, any AnyEvent::MP node will happily accept Storable as framing |
856 |
format, but currently there is no way to make a node use Storable by |
857 |
default (although all nodes will accept it). |
858 |
|
859 |
The default framing protocol is JSON because a) JSON::XS is many times |
860 |
faster for small messages and b) most importantly, after years of |
861 |
experience we found that object serialisation is causing more problems |
862 |
than it solves: Just like function calls, objects simply do not travel |
863 |
easily over the network, mostly because they will always be a copy, so you |
864 |
always have to re-think your design. |
865 |
|
866 |
Keeping your messages simple, concentrating on data structures rather than |
867 |
objects, will keep your messages clean, tidy and efficient. |
868 |
|
869 |
=back |
870 |
|
871 |
=head1 SEE ALSO |
872 |
|
873 |
L<AnyEvent::MP::Intro> - a gentle introduction. |
874 |
|
875 |
L<AnyEvent::MP::Kernel> - more, lower-level, stuff. |
876 |
|
877 |
L<AnyEvent::MP::Global> - network maintainance and port groups, to find |
878 |
your applications. |
879 |
|
880 |
L<AnyEvent::MP::LogCatcher> - simple service to display log messages from |
881 |
all nodes. |
882 |
|
883 |
L<AnyEvent>. |
884 |
|
885 |
=head1 AUTHOR |
886 |
|
887 |
Marc Lehmann <schmorp@schmorp.de> |
888 |
http://home.schmorp.de/ |
889 |
|
890 |
=cut |
891 |
|
892 |
1 |
893 |
|