1 |
=head1 NAME |
2 |
|
3 |
AnyEvent::MP - erlang-style 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 |
# destroy a port again |
36 |
kil $port; # "normal" kill |
37 |
kil $port, my_error => "everything is broken"; # error kill |
38 |
|
39 |
# monitoring |
40 |
mon $localport, $cb->(@msg) # callback is invoked on death |
41 |
mon $localport, $otherport # kill otherport on abnormal death |
42 |
mon $localport, $otherport, @msg # send message on death |
43 |
|
44 |
# temporarily execute code in port context |
45 |
peval $port, sub { die "kill the port!" }; |
46 |
|
47 |
# execute callbacks in $SELF port context |
48 |
my $timer = AE::timer 1, 0, psub { |
49 |
die "kill the port, delayed"; |
50 |
}; |
51 |
|
52 |
=head1 CURRENT STATUS |
53 |
|
54 |
bin/aemp - stable. |
55 |
AnyEvent::MP - stable API, should work. |
56 |
AnyEvent::MP::Intro - explains most concepts. |
57 |
AnyEvent::MP::Kernel - mostly stable API. |
58 |
AnyEvent::MP::Global - stable API. |
59 |
|
60 |
=head1 DESCRIPTION |
61 |
|
62 |
This module (-family) implements a simple message passing framework. |
63 |
|
64 |
Despite its simplicity, you can securely message other processes running |
65 |
on the same or other hosts, and you can supervise entities remotely. |
66 |
|
67 |
For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
68 |
manual page and the examples under F<eg/>. |
69 |
|
70 |
=head1 CONCEPTS |
71 |
|
72 |
=over 4 |
73 |
|
74 |
=item port |
75 |
|
76 |
Not to be confused with a TCP port, a "port" is something you can send |
77 |
messages to (with the C<snd> function). |
78 |
|
79 |
Ports allow you to register C<rcv> handlers that can match all or just |
80 |
some messages. Messages send to ports will not be queued, regardless of |
81 |
anything was listening for them or not. |
82 |
|
83 |
Ports are represented by (printable) strings called "port IDs". |
84 |
|
85 |
=item port ID - C<nodeid#portname> |
86 |
|
87 |
A port ID is the concatenation of a node ID, a hash-mark (C<#>) as |
88 |
separator, and a port name (a printable string of unspecified format). |
89 |
|
90 |
=item node |
91 |
|
92 |
A node is a single process containing at least one port - the node port, |
93 |
which enables nodes to manage each other remotely, and to create new |
94 |
ports. |
95 |
|
96 |
Nodes are either public (have one or more listening ports) or private |
97 |
(no listening ports). Private nodes cannot talk to other private nodes |
98 |
currently, but all nodes can talk to public nodes. |
99 |
|
100 |
Nodes is represented by (printable) strings called "node IDs". |
101 |
|
102 |
=item node ID - C<[A-Za-z0-9_\-.:]*> |
103 |
|
104 |
A node ID is a string that uniquely identifies the node within a |
105 |
network. Depending on the configuration used, node IDs can look like a |
106 |
hostname, a hostname and a port, or a random string. AnyEvent::MP itself |
107 |
doesn't interpret node IDs in any way except to uniquely identify a node. |
108 |
|
109 |
=item binds - C<ip:port> |
110 |
|
111 |
Nodes can only talk to each other by creating some kind of connection to |
112 |
each other. To do this, nodes should listen on one or more local transport |
113 |
endpoints - binds. |
114 |
|
115 |
Currently, only standard C<ip:port> specifications can be used, which |
116 |
specify TCP ports to listen on. So a bind is basically just a tcp socket |
117 |
in listening mode thta accepts conenctions form other nodes. |
118 |
|
119 |
=item seed nodes |
120 |
|
121 |
When a node starts, it knows nothing about the network it is in - it |
122 |
needs to connect to at least one other node that is already in the |
123 |
network. These other nodes are called "seed nodes". |
124 |
|
125 |
Seed nodes themselves are not special - they are seed nodes only because |
126 |
some other node I<uses> them as such, but any node can be used as seed |
127 |
node for other nodes, and eahc node cna use a different set of seed nodes. |
128 |
|
129 |
In addition to discovering the network, seed nodes are also used to |
130 |
maintain the network - all nodes using the same seed node form are part of |
131 |
the same network. If a network is split into multiple subnets because e.g. |
132 |
the network link between the parts goes down, then using the same seed |
133 |
nodes for all nodes ensures that eventually the subnets get merged again. |
134 |
|
135 |
Seed nodes are expected to be long-running, and at least one seed node |
136 |
should always be available. They should also be relatively responsive - a |
137 |
seed node that blocks for long periods will slow down everybody else. |
138 |
|
139 |
For small networks, it's best if every node uses the same set of seed |
140 |
nodes. For large networks, it can be useful to specify "regional" seed |
141 |
nodes for most nodes in an area, and use all seed nodes as seed nodes for |
142 |
each other. What's important is that all seed nodes connections form a |
143 |
complete graph, so that the network cannot split into separate subnets |
144 |
forever. |
145 |
|
146 |
Seed nodes are represented by seed IDs. |
147 |
|
148 |
=item seed IDs - C<host:port> |
149 |
|
150 |
Seed IDs are transport endpoint(s) (usually a hostname/IP address and a |
151 |
TCP port) of nodes that should be used as seed nodes. |
152 |
|
153 |
=item global nodes |
154 |
|
155 |
An AEMP network needs a discovery service - nodes need to know how to |
156 |
connect to other nodes they only know by name. In addition, AEMP offers a |
157 |
distributed "group database", which maps group names to a list of strings |
158 |
- for example, to register worker ports. |
159 |
|
160 |
A network needs at least one global node to work, and allows every node to |
161 |
be a global node. |
162 |
|
163 |
Any node that loads the L<AnyEvent::MP::Global> module becomes a global |
164 |
node and tries to keep connections to all other nodes. So while it can |
165 |
make sense to make every node "global" in small networks, it usually makes |
166 |
sense to only make seed nodes into global nodes in large networks (nodes |
167 |
keep connections to seed nodes and global nodes, so makign them the same |
168 |
reduces overhead). |
169 |
|
170 |
=back |
171 |
|
172 |
=head1 VARIABLES/FUNCTIONS |
173 |
|
174 |
=over 4 |
175 |
|
176 |
=cut |
177 |
|
178 |
package AnyEvent::MP; |
179 |
|
180 |
use AnyEvent::MP::Kernel; |
181 |
|
182 |
use common::sense; |
183 |
|
184 |
use Carp (); |
185 |
|
186 |
use AE (); |
187 |
|
188 |
use base "Exporter"; |
189 |
|
190 |
our $VERSION = '1.30'; |
191 |
|
192 |
our @EXPORT = qw( |
193 |
NODE $NODE *SELF node_of after |
194 |
configure |
195 |
snd rcv mon mon_guard kil psub peval spawn cal |
196 |
port |
197 |
); |
198 |
|
199 |
our $SELF; |
200 |
|
201 |
sub _self_die() { |
202 |
my $msg = $@; |
203 |
$msg =~ s/\n+$// unless ref $msg; |
204 |
kil $SELF, die => $msg; |
205 |
} |
206 |
|
207 |
=item $thisnode = NODE / $NODE |
208 |
|
209 |
The C<NODE> function returns, and the C<$NODE> variable contains, the node |
210 |
ID of the node running in the current process. This value is initialised by |
211 |
a call to C<configure>. |
212 |
|
213 |
=item $nodeid = node_of $port |
214 |
|
215 |
Extracts and returns the node ID from a port ID or a node ID. |
216 |
|
217 |
=item configure $profile, key => value... |
218 |
|
219 |
=item configure key => value... |
220 |
|
221 |
Before a node can talk to other nodes on the network (i.e. enter |
222 |
"distributed mode") it has to configure itself - the minimum a node needs |
223 |
to know is its own name, and optionally it should know the addresses of |
224 |
some other nodes in the network to discover other nodes. |
225 |
|
226 |
The key/value pairs are basically the same ones as documented for the |
227 |
F<aemp> command line utility (sans the set/del prefix). |
228 |
|
229 |
This function configures a node - it must be called exactly once (or |
230 |
never) before calling other AnyEvent::MP functions. |
231 |
|
232 |
=over 4 |
233 |
|
234 |
=item step 1, gathering configuration from profiles |
235 |
|
236 |
The function first looks up a profile in the aemp configuration (see the |
237 |
L<aemp> commandline utility). The profile name can be specified via the |
238 |
named C<profile> parameter or can simply be the first parameter). If it is |
239 |
missing, then the nodename (F<uname -n>) will be used as profile name. |
240 |
|
241 |
The profile data is then gathered as follows: |
242 |
|
243 |
First, all remaining key => value pairs (all of which are conveniently |
244 |
undocumented at the moment) will be interpreted as configuration |
245 |
data. Then they will be overwritten by any values specified in the global |
246 |
default configuration (see the F<aemp> utility), then the chain of |
247 |
profiles chosen by the profile name (and any C<parent> attributes). |
248 |
|
249 |
That means that the values specified in the profile have highest priority |
250 |
and the values specified directly via C<configure> have lowest priority, |
251 |
and can only be used to specify defaults. |
252 |
|
253 |
If the profile specifies a node ID, then this will become the node ID of |
254 |
this process. If not, then the profile name will be used as node ID. The |
255 |
special node ID of C<anon/> will be replaced by a random node ID. |
256 |
|
257 |
=item step 2, bind listener sockets |
258 |
|
259 |
The next step is to look up the binds in the profile, followed by binding |
260 |
aemp protocol listeners on all binds specified (it is possible and valid |
261 |
to have no binds, meaning that the node cannot be contacted form the |
262 |
outside. This means the node cannot talk to other nodes that also have no |
263 |
binds, but it can still talk to all "normal" nodes). |
264 |
|
265 |
If the profile does not specify a binds list, then a default of C<*> is |
266 |
used, meaning the node will bind on a dynamically-assigned port on every |
267 |
local IP address it finds. |
268 |
|
269 |
=item step 3, connect to seed nodes |
270 |
|
271 |
As the last step, the seed ID list from the profile is passed to the |
272 |
L<AnyEvent::MP::Global> module, which will then use it to keep |
273 |
connectivity with at least one node at any point in time. |
274 |
|
275 |
=back |
276 |
|
277 |
Example: become a distributed node using the local node name as profile. |
278 |
This should be the most common form of invocation for "daemon"-type nodes. |
279 |
|
280 |
configure |
281 |
|
282 |
Example: become an anonymous node. This form is often used for commandline |
283 |
clients. |
284 |
|
285 |
configure nodeid => "anon/"; |
286 |
|
287 |
Example: configure a node using a profile called seed, which si suitable |
288 |
for a seed node as it binds on all local addresses on a fixed port (4040, |
289 |
customary for aemp). |
290 |
|
291 |
# use the aemp commandline utility |
292 |
# aemp profile seed nodeid anon/ binds '*:4040' |
293 |
|
294 |
# then use it |
295 |
configure profile => "seed"; |
296 |
|
297 |
# or simply use aemp from the shell again: |
298 |
# aemp run profile seed |
299 |
|
300 |
# or provide a nicer-to-remember nodeid |
301 |
# aemp run profile seed nodeid "$(hostname)" |
302 |
|
303 |
=item $SELF |
304 |
|
305 |
Contains the current port id while executing C<rcv> callbacks or C<psub> |
306 |
blocks. |
307 |
|
308 |
=item *SELF, SELF, %SELF, @SELF... |
309 |
|
310 |
Due to some quirks in how perl exports variables, it is impossible to |
311 |
just export C<$SELF>, all the symbols named C<SELF> are exported by this |
312 |
module, but only C<$SELF> is currently used. |
313 |
|
314 |
=item snd $port, type => @data |
315 |
|
316 |
=item snd $port, @msg |
317 |
|
318 |
Send the given message to the given port, which can identify either a |
319 |
local or a remote port, and must be a port ID. |
320 |
|
321 |
While the message can be almost anything, it is highly recommended to |
322 |
use a string as first element (a port ID, or some word that indicates a |
323 |
request type etc.) and to consist if only simple perl values (scalars, |
324 |
arrays, hashes) - if you think you need to pass an object, think again. |
325 |
|
326 |
The message data logically becomes read-only after a call to this |
327 |
function: modifying any argument (or values referenced by them) is |
328 |
forbidden, as there can be considerable time between the call to C<snd> |
329 |
and the time the message is actually being serialised - in fact, it might |
330 |
never be copied as within the same process it is simply handed to the |
331 |
receiving port. |
332 |
|
333 |
The type of data you can transfer depends on the transport protocol: when |
334 |
JSON is used, then only strings, numbers and arrays and hashes consisting |
335 |
of those are allowed (no objects). When Storable is used, then anything |
336 |
that Storable can serialise and deserialise is allowed, and for the local |
337 |
node, anything can be passed. Best rely only on the common denominator of |
338 |
these. |
339 |
|
340 |
=item $local_port = port |
341 |
|
342 |
Create a new local port object and returns its port ID. Initially it has |
343 |
no callbacks set and will throw an error when it receives messages. |
344 |
|
345 |
=item $local_port = port { my @msg = @_ } |
346 |
|
347 |
Creates a new local port, and returns its ID. Semantically the same as |
348 |
creating a port and calling C<rcv $port, $callback> on it. |
349 |
|
350 |
The block will be called for every message received on the port, with the |
351 |
global variable C<$SELF> set to the port ID. Runtime errors will cause the |
352 |
port to be C<kil>ed. The message will be passed as-is, no extra argument |
353 |
(i.e. no port ID) will be passed to the callback. |
354 |
|
355 |
If you want to stop/destroy the port, simply C<kil> it: |
356 |
|
357 |
my $port = port { |
358 |
my @msg = @_; |
359 |
... |
360 |
kil $SELF; |
361 |
}; |
362 |
|
363 |
=cut |
364 |
|
365 |
sub rcv($@); |
366 |
|
367 |
sub _kilme { |
368 |
die "received message on port without callback"; |
369 |
} |
370 |
|
371 |
sub port(;&) { |
372 |
my $id = "$UNIQ." . $ID++; |
373 |
my $port = "$NODE#$id"; |
374 |
|
375 |
rcv $port, shift || \&_kilme; |
376 |
|
377 |
$port |
378 |
} |
379 |
|
380 |
=item rcv $local_port, $callback->(@msg) |
381 |
|
382 |
Replaces the default callback on the specified port. There is no way to |
383 |
remove the default callback: use C<sub { }> to disable it, or better |
384 |
C<kil> the port when it is no longer needed. |
385 |
|
386 |
The global C<$SELF> (exported by this module) contains C<$port> while |
387 |
executing the callback. Runtime errors during callback execution will |
388 |
result in the port being C<kil>ed. |
389 |
|
390 |
The default callback received all messages not matched by a more specific |
391 |
C<tag> match. |
392 |
|
393 |
=item rcv $local_port, tag => $callback->(@msg_without_tag), ... |
394 |
|
395 |
Register (or replace) callbacks to be called on messages starting with the |
396 |
given tag on the given port (and return the port), or unregister it (when |
397 |
C<$callback> is C<$undef> or missing). There can only be one callback |
398 |
registered for each tag. |
399 |
|
400 |
The original message will be passed to the callback, after the first |
401 |
element (the tag) has been removed. The callback will use the same |
402 |
environment as the default callback (see above). |
403 |
|
404 |
Example: create a port and bind receivers on it in one go. |
405 |
|
406 |
my $port = rcv port, |
407 |
msg1 => sub { ... }, |
408 |
msg2 => sub { ... }, |
409 |
; |
410 |
|
411 |
Example: create a port, bind receivers and send it in a message elsewhere |
412 |
in one go: |
413 |
|
414 |
snd $otherport, reply => |
415 |
rcv port, |
416 |
msg1 => sub { ... }, |
417 |
... |
418 |
; |
419 |
|
420 |
Example: temporarily register a rcv callback for a tag matching some port |
421 |
(e.g. for an rpc reply) and unregister it after a message was received. |
422 |
|
423 |
rcv $port, $otherport => sub { |
424 |
my @reply = @_; |
425 |
|
426 |
rcv $SELF, $otherport; |
427 |
}; |
428 |
|
429 |
=cut |
430 |
|
431 |
sub rcv($@) { |
432 |
my $port = shift; |
433 |
my ($nodeid, $portid) = split /#/, $port, 2; |
434 |
|
435 |
$NODE{$nodeid} == $NODE{""} |
436 |
or Carp::croak "$port: rcv can only be called on local ports, caught"; |
437 |
|
438 |
while (@_) { |
439 |
if (ref $_[0]) { |
440 |
if (my $self = $PORT_DATA{$portid}) { |
441 |
"AnyEvent::MP::Port" eq ref $self |
442 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
443 |
|
444 |
$self->[0] = shift; |
445 |
} else { |
446 |
my $cb = shift; |
447 |
$PORT{$portid} = sub { |
448 |
local $SELF = $port; |
449 |
eval { &$cb }; _self_die if $@; |
450 |
}; |
451 |
} |
452 |
} elsif (defined $_[0]) { |
453 |
my $self = $PORT_DATA{$portid} ||= do { |
454 |
my $self = bless [$PORT{$portid} || sub { }, { }, $port], "AnyEvent::MP::Port"; |
455 |
|
456 |
$PORT{$portid} = sub { |
457 |
local $SELF = $port; |
458 |
|
459 |
if (my $cb = $self->[1]{$_[0]}) { |
460 |
shift; |
461 |
eval { &$cb }; _self_die if $@; |
462 |
} else { |
463 |
&{ $self->[0] }; |
464 |
} |
465 |
}; |
466 |
|
467 |
$self |
468 |
}; |
469 |
|
470 |
"AnyEvent::MP::Port" eq ref $self |
471 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
472 |
|
473 |
my ($tag, $cb) = splice @_, 0, 2; |
474 |
|
475 |
if (defined $cb) { |
476 |
$self->[1]{$tag} = $cb; |
477 |
} else { |
478 |
delete $self->[1]{$tag}; |
479 |
} |
480 |
} |
481 |
} |
482 |
|
483 |
$port |
484 |
} |
485 |
|
486 |
=item peval $port, $coderef[, @args] |
487 |
|
488 |
Evaluates the given C<$codref> within the contetx of C<$port>, that is, |
489 |
when the code throews an exception the C<$port> will be killed. |
490 |
|
491 |
Any remaining args will be passed to the callback. Any return values will |
492 |
be returned to the caller. |
493 |
|
494 |
This is useful when you temporarily want to execute code in the context of |
495 |
a port. |
496 |
|
497 |
Example: create a port and run some initialisation code in it's context. |
498 |
|
499 |
my $port = port { ... }; |
500 |
|
501 |
peval $port, sub { |
502 |
init |
503 |
or die "unable to init"; |
504 |
}; |
505 |
|
506 |
=cut |
507 |
|
508 |
sub peval($$) { |
509 |
local $SELF = shift; |
510 |
my $cb = shift; |
511 |
|
512 |
if (wantarray) { |
513 |
my @res = eval { &$cb }; |
514 |
_self_die if $@; |
515 |
@res |
516 |
} else { |
517 |
my $res = eval { &$cb }; |
518 |
_self_die if $@; |
519 |
$res |
520 |
} |
521 |
} |
522 |
|
523 |
=item $closure = psub { BLOCK } |
524 |
|
525 |
Remembers C<$SELF> and creates a closure out of the BLOCK. When the |
526 |
closure is executed, sets up the environment in the same way as in C<rcv> |
527 |
callbacks, i.e. runtime errors will cause the port to get C<kil>ed. |
528 |
|
529 |
The effect is basically as if it returned C<< sub { peval $SELF, sub { |
530 |
BLOCK }, @_ } >>. |
531 |
|
532 |
This is useful when you register callbacks from C<rcv> callbacks: |
533 |
|
534 |
rcv delayed_reply => sub { |
535 |
my ($delay, @reply) = @_; |
536 |
my $timer = AE::timer $delay, 0, psub { |
537 |
snd @reply, $SELF; |
538 |
}; |
539 |
}; |
540 |
|
541 |
=cut |
542 |
|
543 |
sub psub(&) { |
544 |
my $cb = shift; |
545 |
|
546 |
my $port = $SELF |
547 |
or Carp::croak "psub can only be called from within rcv or psub callbacks, not"; |
548 |
|
549 |
sub { |
550 |
local $SELF = $port; |
551 |
|
552 |
if (wantarray) { |
553 |
my @res = eval { &$cb }; |
554 |
_self_die if $@; |
555 |
@res |
556 |
} else { |
557 |
my $res = eval { &$cb }; |
558 |
_self_die if $@; |
559 |
$res |
560 |
} |
561 |
} |
562 |
} |
563 |
|
564 |
=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies |
565 |
|
566 |
=item $guard = mon $port, $rcvport # kill $rcvport when $port dies |
567 |
|
568 |
=item $guard = mon $port # kill $SELF when $port dies |
569 |
|
570 |
=item $guard = mon $port, $rcvport, @msg # send a message when $port dies |
571 |
|
572 |
Monitor the given port and do something when the port is killed or |
573 |
messages to it were lost, and optionally return a guard that can be used |
574 |
to stop monitoring again. |
575 |
|
576 |
In the first form (callback), the callback is simply called with any |
577 |
number of C<@reason> elements (no @reason means that the port was deleted |
578 |
"normally"). Note also that I<< the callback B<must> never die >>, so use |
579 |
C<eval> if unsure. |
580 |
|
581 |
In the second form (another port given), the other port (C<$rcvport>) |
582 |
will be C<kil>'ed with C<@reason>, if a @reason was specified, i.e. on |
583 |
"normal" kils nothing happens, while under all other conditions, the other |
584 |
port is killed with the same reason. |
585 |
|
586 |
The third form (kill self) is the same as the second form, except that |
587 |
C<$rvport> defaults to C<$SELF>. |
588 |
|
589 |
In the last form (message), a message of the form C<@msg, @reason> will be |
590 |
C<snd>. |
591 |
|
592 |
Monitoring-actions are one-shot: once messages are lost (and a monitoring |
593 |
alert was raised), they are removed and will not trigger again. |
594 |
|
595 |
As a rule of thumb, monitoring requests should always monitor a port from |
596 |
a local port (or callback). The reason is that kill messages might get |
597 |
lost, just like any other message. Another less obvious reason is that |
598 |
even monitoring requests can get lost (for example, when the connection |
599 |
to the other node goes down permanently). When monitoring a port locally |
600 |
these problems do not exist. |
601 |
|
602 |
C<mon> effectively guarantees that, in the absence of hardware failures, |
603 |
after starting the monitor, either all messages sent to the port will |
604 |
arrive, or the monitoring action will be invoked after possible message |
605 |
loss has been detected. No messages will be lost "in between" (after |
606 |
the first lost message no further messages will be received by the |
607 |
port). After the monitoring action was invoked, further messages might get |
608 |
delivered again. |
609 |
|
610 |
Inter-host-connection timeouts and monitoring depend on the transport |
611 |
used. The only transport currently implemented is TCP, and AnyEvent::MP |
612 |
relies on TCP to detect node-downs (this can take 10-15 minutes on a |
613 |
non-idle connection, and usually around two hours for idle connections). |
614 |
|
615 |
This means that monitoring is good for program errors and cleaning up |
616 |
stuff eventually, but they are no replacement for a timeout when you need |
617 |
to ensure some maximum latency. |
618 |
|
619 |
Example: call a given callback when C<$port> is killed. |
620 |
|
621 |
mon $port, sub { warn "port died because of <@_>\n" }; |
622 |
|
623 |
Example: kill ourselves when C<$port> is killed abnormally. |
624 |
|
625 |
mon $port; |
626 |
|
627 |
Example: send us a restart message when another C<$port> is killed. |
628 |
|
629 |
mon $port, $self => "restart"; |
630 |
|
631 |
=cut |
632 |
|
633 |
sub mon { |
634 |
my ($nodeid, $port) = split /#/, shift, 2; |
635 |
|
636 |
my $node = $NODE{$nodeid} || add_node $nodeid; |
637 |
|
638 |
my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,'; |
639 |
|
640 |
unless (ref $cb) { |
641 |
if (@_) { |
642 |
# send a kill info message |
643 |
my (@msg) = ($cb, @_); |
644 |
$cb = sub { snd @msg, @_ }; |
645 |
} else { |
646 |
# simply kill other port |
647 |
my $port = $cb; |
648 |
$cb = sub { kil $port, @_ if @_ }; |
649 |
} |
650 |
} |
651 |
|
652 |
$node->monitor ($port, $cb); |
653 |
|
654 |
defined wantarray |
655 |
and ($cb += 0, AnyEvent::Util::guard { $node->unmonitor ($port, $cb) }) |
656 |
} |
657 |
|
658 |
=item $guard = mon_guard $port, $ref, $ref... |
659 |
|
660 |
Monitors the given C<$port> and keeps the passed references. When the port |
661 |
is killed, the references will be freed. |
662 |
|
663 |
Optionally returns a guard that will stop the monitoring. |
664 |
|
665 |
This function is useful when you create e.g. timers or other watchers and |
666 |
want to free them when the port gets killed (note the use of C<psub>): |
667 |
|
668 |
$port->rcv (start => sub { |
669 |
my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub { |
670 |
undef $timer if 0.9 < rand; |
671 |
}); |
672 |
}); |
673 |
|
674 |
=cut |
675 |
|
676 |
sub mon_guard { |
677 |
my ($port, @refs) = @_; |
678 |
|
679 |
#TODO: mon-less form? |
680 |
|
681 |
mon $port, sub { 0 && @refs } |
682 |
} |
683 |
|
684 |
=item kil $port[, @reason] |
685 |
|
686 |
Kill the specified port with the given C<@reason>. |
687 |
|
688 |
If no C<@reason> is specified, then the port is killed "normally" - |
689 |
monitor callback will be invoked, but the kil will not cause linked ports |
690 |
(C<mon $mport, $lport> form) to get killed. |
691 |
|
692 |
If a C<@reason> is specified, then linked ports (C<mon $mport, $lport> |
693 |
form) get killed with the same reason. |
694 |
|
695 |
Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks |
696 |
will be reported as reason C<< die => $@ >>. |
697 |
|
698 |
Transport/communication errors are reported as C<< transport_error => |
699 |
$message >>. |
700 |
|
701 |
=cut |
702 |
|
703 |
=item $port = spawn $node, $initfunc[, @initdata] |
704 |
|
705 |
Creates a port on the node C<$node> (which can also be a port ID, in which |
706 |
case it's the node where that port resides). |
707 |
|
708 |
The port ID of the newly created port is returned immediately, and it is |
709 |
possible to immediately start sending messages or to monitor the port. |
710 |
|
711 |
After the port has been created, the init function is called on the remote |
712 |
node, in the same context as a C<rcv> callback. This function must be a |
713 |
fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To |
714 |
specify a function in the main program, use C<::name>. |
715 |
|
716 |
If the function doesn't exist, then the node tries to C<require> |
717 |
the package, then the package above the package and so on (e.g. |
718 |
C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
719 |
exists or it runs out of package names. |
720 |
|
721 |
The init function is then called with the newly-created port as context |
722 |
object (C<$SELF>) and the C<@initdata> values as arguments. It I<must> |
723 |
call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise |
724 |
the port might not get created. |
725 |
|
726 |
A common idiom is to pass a local port, immediately monitor the spawned |
727 |
port, and in the remote init function, immediately monitor the passed |
728 |
local port. This two-way monitoring ensures that both ports get cleaned up |
729 |
when there is a problem. |
730 |
|
731 |
C<spawn> guarantees that the C<$initfunc> has no visible effects on the |
732 |
caller before C<spawn> returns (by delaying invocation when spawn is |
733 |
called for the local node). |
734 |
|
735 |
Example: spawn a chat server port on C<$othernode>. |
736 |
|
737 |
# this node, executed from within a port context: |
738 |
my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
739 |
mon $server; |
740 |
|
741 |
# init function on C<$othernode> |
742 |
sub connect { |
743 |
my ($srcport) = @_; |
744 |
|
745 |
mon $srcport; |
746 |
|
747 |
rcv $SELF, sub { |
748 |
... |
749 |
}; |
750 |
} |
751 |
|
752 |
=cut |
753 |
|
754 |
sub _spawn { |
755 |
my $port = shift; |
756 |
my $init = shift; |
757 |
|
758 |
# rcv will create the actual port |
759 |
local $SELF = "$NODE#$port"; |
760 |
eval { |
761 |
&{ load_func $init } |
762 |
}; |
763 |
_self_die if $@; |
764 |
} |
765 |
|
766 |
sub spawn(@) { |
767 |
my ($nodeid, undef) = split /#/, shift, 2; |
768 |
|
769 |
my $id = "$RUNIQ." . $ID++; |
770 |
|
771 |
$_[0] =~ /::/ |
772 |
or Carp::croak "spawn init function must be a fully-qualified name, caught"; |
773 |
|
774 |
snd_to_func $nodeid, "AnyEvent::MP::_spawn" => $id, @_; |
775 |
|
776 |
"$nodeid#$id" |
777 |
} |
778 |
|
779 |
=item after $timeout, @msg |
780 |
|
781 |
=item after $timeout, $callback |
782 |
|
783 |
Either sends the given message, or call the given callback, after the |
784 |
specified number of seconds. |
785 |
|
786 |
This is simply a utility function that comes in handy at times - the |
787 |
AnyEvent::MP author is not convinced of the wisdom of having it, though, |
788 |
so it may go away in the future. |
789 |
|
790 |
=cut |
791 |
|
792 |
sub after($@) { |
793 |
my ($timeout, @action) = @_; |
794 |
|
795 |
my $t; $t = AE::timer $timeout, 0, sub { |
796 |
undef $t; |
797 |
ref $action[0] |
798 |
? $action[0]() |
799 |
: snd @action; |
800 |
}; |
801 |
} |
802 |
|
803 |
=item cal $port, @msg, $callback[, $timeout] |
804 |
|
805 |
A simple form of RPC - sends a message to the given C<$port> with the |
806 |
given contents (C<@msg>), but adds a reply port to the message. |
807 |
|
808 |
The reply port is created temporarily just for the purpose of receiving |
809 |
the reply, and will be C<kil>ed when no longer needed. |
810 |
|
811 |
A reply message sent to the port is passed to the C<$callback> as-is. |
812 |
|
813 |
If an optional time-out (in seconds) is given and it is not C<undef>, |
814 |
then the callback will be called without any arguments after the time-out |
815 |
elapsed and the port is C<kil>ed. |
816 |
|
817 |
If no time-out is given (or it is C<undef>), then the local port will |
818 |
monitor the remote port instead, so it eventually gets cleaned-up. |
819 |
|
820 |
Currently this function returns the temporary port, but this "feature" |
821 |
might go in future versions unless you can make a convincing case that |
822 |
this is indeed useful for something. |
823 |
|
824 |
=cut |
825 |
|
826 |
sub cal(@) { |
827 |
my $timeout = ref $_[-1] ? undef : pop; |
828 |
my $cb = pop; |
829 |
|
830 |
my $port = port { |
831 |
undef $timeout; |
832 |
kil $SELF; |
833 |
&$cb; |
834 |
}; |
835 |
|
836 |
if (defined $timeout) { |
837 |
$timeout = AE::timer $timeout, 0, sub { |
838 |
undef $timeout; |
839 |
kil $port; |
840 |
$cb->(); |
841 |
}; |
842 |
} else { |
843 |
mon $_[0], sub { |
844 |
kil $port; |
845 |
$cb->(); |
846 |
}; |
847 |
} |
848 |
|
849 |
push @_, $port; |
850 |
&snd; |
851 |
|
852 |
$port |
853 |
} |
854 |
|
855 |
=back |
856 |
|
857 |
=head1 AnyEvent::MP vs. Distributed Erlang |
858 |
|
859 |
AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
860 |
== aemp node, Erlang process == aemp port), so many of the documents and |
861 |
programming techniques employed by Erlang apply to AnyEvent::MP. Here is a |
862 |
sample: |
863 |
|
864 |
http://www.erlang.se/doc/programming_rules.shtml |
865 |
http://erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4 |
866 |
http://erlang.org/download/erlang-book-part1.pdf # chapters 5 and 6 |
867 |
http://erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5 |
868 |
|
869 |
Despite the similarities, there are also some important differences: |
870 |
|
871 |
=over 4 |
872 |
|
873 |
=item * Node IDs are arbitrary strings in AEMP. |
874 |
|
875 |
Erlang relies on special naming and DNS to work everywhere in the same |
876 |
way. AEMP relies on each node somehow knowing its own address(es) (e.g. by |
877 |
configuration or DNS), and possibly the addresses of some seed nodes, but |
878 |
will otherwise discover other nodes (and their IDs) itself. |
879 |
|
880 |
=item * Erlang has a "remote ports are like local ports" philosophy, AEMP |
881 |
uses "local ports are like remote ports". |
882 |
|
883 |
The failure modes for local ports are quite different (runtime errors |
884 |
only) then for remote ports - when a local port dies, you I<know> it dies, |
885 |
when a connection to another node dies, you know nothing about the other |
886 |
port. |
887 |
|
888 |
Erlang pretends remote ports are as reliable as local ports, even when |
889 |
they are not. |
890 |
|
891 |
AEMP encourages a "treat remote ports differently" philosophy, with local |
892 |
ports being the special case/exception, where transport errors cannot |
893 |
occur. |
894 |
|
895 |
=item * Erlang uses processes and a mailbox, AEMP does not queue. |
896 |
|
897 |
Erlang uses processes that selectively receive messages out of order, and |
898 |
therefore needs a queue. AEMP is event based, queuing messages would serve |
899 |
no useful purpose. For the same reason the pattern-matching abilities |
900 |
of AnyEvent::MP are more limited, as there is little need to be able to |
901 |
filter messages without dequeuing them. |
902 |
|
903 |
This is not a philosophical difference, but simply stems from AnyEvent::MP |
904 |
being event-based, while Erlang is process-based. |
905 |
|
906 |
You cna have a look at L<Coro::MP> for a more Erlang-like process model on |
907 |
top of AEMP and Coro threads. |
908 |
|
909 |
=item * Erlang sends are synchronous, AEMP sends are asynchronous. |
910 |
|
911 |
Sending messages in Erlang is synchronous and blocks the process until |
912 |
a conenction has been established and the message sent (and so does not |
913 |
need a queue that can overflow). AEMP sends return immediately, connection |
914 |
establishment is handled in the background. |
915 |
|
916 |
=item * Erlang suffers from silent message loss, AEMP does not. |
917 |
|
918 |
Erlang implements few guarantees on messages delivery - messages can get |
919 |
lost without any of the processes realising it (i.e. you send messages a, |
920 |
b, and c, and the other side only receives messages a and c). |
921 |
|
922 |
AEMP guarantees (modulo hardware errors) correct ordering, and the |
923 |
guarantee that after one message is lost, all following ones sent to the |
924 |
same port are lost as well, until monitoring raises an error, so there are |
925 |
no silent "holes" in the message sequence. |
926 |
|
927 |
If you want your software to be very reliable, you have to cope with |
928 |
corrupted and even out-of-order messages in both Erlang and AEMP. AEMP |
929 |
simply tries to work better in common error cases, such as when a network |
930 |
link goes down. |
931 |
|
932 |
=item * Erlang can send messages to the wrong port, AEMP does not. |
933 |
|
934 |
In Erlang it is quite likely that a node that restarts reuses an Erlang |
935 |
process ID known to other nodes for a completely different process, |
936 |
causing messages destined for that process to end up in an unrelated |
937 |
process. |
938 |
|
939 |
AEMP does not reuse port IDs, so old messages or old port IDs floating |
940 |
around in the network will not be sent to an unrelated port. |
941 |
|
942 |
=item * Erlang uses unprotected connections, AEMP uses secure |
943 |
authentication and can use TLS. |
944 |
|
945 |
AEMP can use a proven protocol - TLS - to protect connections and |
946 |
securely authenticate nodes. |
947 |
|
948 |
=item * The AEMP protocol is optimised for both text-based and binary |
949 |
communications. |
950 |
|
951 |
The AEMP protocol, unlike the Erlang protocol, supports both programming |
952 |
language independent text-only protocols (good for debugging), and binary, |
953 |
language-specific serialisers (e.g. Storable). By default, unless TLS is |
954 |
used, the protocol is actually completely text-based. |
955 |
|
956 |
It has also been carefully designed to be implementable in other languages |
957 |
with a minimum of work while gracefully degrading functionality to make the |
958 |
protocol simple. |
959 |
|
960 |
=item * AEMP has more flexible monitoring options than Erlang. |
961 |
|
962 |
In Erlang, you can chose to receive I<all> exit signals as messages or |
963 |
I<none>, there is no in-between, so monitoring single Erlang processes is |
964 |
difficult to implement. |
965 |
|
966 |
Monitoring in AEMP is more flexible than in Erlang, as one can choose |
967 |
between automatic kill, exit message or callback on a per-port basis. |
968 |
|
969 |
=item * Erlang tries to hide remote/local connections, AEMP does not. |
970 |
|
971 |
Monitoring in Erlang is not an indicator of process death/crashes, in the |
972 |
same way as linking is (except linking is unreliable in Erlang). |
973 |
|
974 |
In AEMP, you don't "look up" registered port names or send to named ports |
975 |
that might or might not be persistent. Instead, you normally spawn a port |
976 |
on the remote node. The init function monitors you, and you monitor the |
977 |
remote port. Since both monitors are local to the node, they are much more |
978 |
reliable (no need for C<spawn_link>). |
979 |
|
980 |
This also saves round-trips and avoids sending messages to the wrong port |
981 |
(hard to do in Erlang). |
982 |
|
983 |
=back |
984 |
|
985 |
=head1 RATIONALE |
986 |
|
987 |
=over 4 |
988 |
|
989 |
=item Why strings for port and node IDs, why not objects? |
990 |
|
991 |
We considered "objects", but found that the actual number of methods |
992 |
that can be called are quite low. Since port and node IDs travel over |
993 |
the network frequently, the serialising/deserialising would add lots of |
994 |
overhead, as well as having to keep a proxy object everywhere. |
995 |
|
996 |
Strings can easily be printed, easily serialised etc. and need no special |
997 |
procedures to be "valid". |
998 |
|
999 |
And as a result, a port with just a default receiver consists of a single |
1000 |
code reference stored in a global hash - it can't become much cheaper. |
1001 |
|
1002 |
=item Why favour JSON, why not a real serialising format such as Storable? |
1003 |
|
1004 |
In fact, any AnyEvent::MP node will happily accept Storable as framing |
1005 |
format, but currently there is no way to make a node use Storable by |
1006 |
default (although all nodes will accept it). |
1007 |
|
1008 |
The default framing protocol is JSON because a) JSON::XS is many times |
1009 |
faster for small messages and b) most importantly, after years of |
1010 |
experience we found that object serialisation is causing more problems |
1011 |
than it solves: Just like function calls, objects simply do not travel |
1012 |
easily over the network, mostly because they will always be a copy, so you |
1013 |
always have to re-think your design. |
1014 |
|
1015 |
Keeping your messages simple, concentrating on data structures rather than |
1016 |
objects, will keep your messages clean, tidy and efficient. |
1017 |
|
1018 |
=back |
1019 |
|
1020 |
=head1 SEE ALSO |
1021 |
|
1022 |
L<AnyEvent::MP::Intro> - a gentle introduction. |
1023 |
|
1024 |
L<AnyEvent::MP::Kernel> - more, lower-level, stuff. |
1025 |
|
1026 |
L<AnyEvent::MP::Global> - network maintenance and port groups, to find |
1027 |
your applications. |
1028 |
|
1029 |
L<AnyEvent::MP::DataConn> - establish data connections between nodes. |
1030 |
|
1031 |
L<AnyEvent::MP::LogCatcher> - simple service to display log messages from |
1032 |
all nodes. |
1033 |
|
1034 |
L<AnyEvent>. |
1035 |
|
1036 |
=head1 AUTHOR |
1037 |
|
1038 |
Marc Lehmann <schmorp@schmorp.de> |
1039 |
http://home.schmorp.de/ |
1040 |
|
1041 |
=cut |
1042 |
|
1043 |
1 |
1044 |
|