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 $port, $cb->(@msg) # callback is invoked on death |
41 |
mon $port, $localport # kill localport on abnormal death |
42 |
mon $port, $localport, @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 |
# distributed database - modification |
53 |
db_set $family => $subkey [=> $value] # add a subkey |
54 |
db_del $family => $subkey... # delete one or more subkeys |
55 |
db_reg $family => $port [=> $value] # register a port |
56 |
|
57 |
# distributed database - queries |
58 |
db_family $family => $cb->(\%familyhash) |
59 |
db_keys $family => $cb->(\@keys) |
60 |
db_values $family => $cb->(\@values) |
61 |
|
62 |
# distributed database - monitoring a family |
63 |
db_mon $family => $cb->(\%familyhash, \@added, \@changed, \@deleted) |
64 |
|
65 |
=head1 DESCRIPTION |
66 |
|
67 |
This module (-family) implements a simple message passing framework. |
68 |
|
69 |
Despite its simplicity, you can securely message other processes running |
70 |
on the same or other hosts, and you can supervise entities remotely. |
71 |
|
72 |
For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
73 |
manual page and the examples under F<eg/>. |
74 |
|
75 |
=head1 CONCEPTS |
76 |
|
77 |
=over 4 |
78 |
|
79 |
=item port |
80 |
|
81 |
Not to be confused with a TCP port, a "port" is something you can send |
82 |
messages to (with the C<snd> function). |
83 |
|
84 |
Ports allow you to register C<rcv> handlers that can match all or just |
85 |
some messages. Messages send to ports will not be queued, regardless of |
86 |
anything was listening for them or not. |
87 |
|
88 |
Ports are represented by (printable) strings called "port IDs". |
89 |
|
90 |
=item port ID - C<nodeid#portname> |
91 |
|
92 |
A port ID is the concatenation of a node ID, a hash-mark (C<#>) |
93 |
as separator, and a port name (a printable string of unspecified |
94 |
format created by AnyEvent::MP). |
95 |
|
96 |
=item node |
97 |
|
98 |
A node is a single process containing at least one port - the node port, |
99 |
which enables nodes to manage each other remotely, and to create new |
100 |
ports. |
101 |
|
102 |
Nodes are either public (have one or more listening ports) or private |
103 |
(no listening ports). Private nodes cannot talk to other private nodes |
104 |
currently, but all nodes can talk to public nodes. |
105 |
|
106 |
Nodes is represented by (printable) strings called "node IDs". |
107 |
|
108 |
=item node ID - C<[A-Za-z0-9_\-.:]*> |
109 |
|
110 |
A node ID is a string that uniquely identifies the node within a |
111 |
network. Depending on the configuration used, node IDs can look like a |
112 |
hostname, a hostname and a port, or a random string. AnyEvent::MP itself |
113 |
doesn't interpret node IDs in any way except to uniquely identify a node. |
114 |
|
115 |
=item binds - C<ip:port> |
116 |
|
117 |
Nodes can only talk to each other by creating some kind of connection to |
118 |
each other. To do this, nodes should listen on one or more local transport |
119 |
endpoints - binds. |
120 |
|
121 |
Currently, only standard C<ip:port> specifications can be used, which |
122 |
specify TCP ports to listen on. So a bind is basically just a tcp socket |
123 |
in listening mode thta accepts conenctions form other nodes. |
124 |
|
125 |
=item seed nodes |
126 |
|
127 |
When a node starts, it knows nothing about the network it is in - it |
128 |
needs to connect to at least one other node that is already in the |
129 |
network. These other nodes are called "seed nodes". |
130 |
|
131 |
Seed nodes themselves are not special - they are seed nodes only because |
132 |
some other node I<uses> them as such, but any node can be used as seed |
133 |
node for other nodes, and eahc node cna use a different set of seed nodes. |
134 |
|
135 |
In addition to discovering the network, seed nodes are also used to |
136 |
maintain the network - all nodes using the same seed node form are part of |
137 |
the same network. If a network is split into multiple subnets because e.g. |
138 |
the network link between the parts goes down, then using the same seed |
139 |
nodes for all nodes ensures that eventually the subnets get merged again. |
140 |
|
141 |
Seed nodes are expected to be long-running, and at least one seed node |
142 |
should always be available. They should also be relatively responsive - a |
143 |
seed node that blocks for long periods will slow down everybody else. |
144 |
|
145 |
For small networks, it's best if every node uses the same set of seed |
146 |
nodes. For large networks, it can be useful to specify "regional" seed |
147 |
nodes for most nodes in an area, and use all seed nodes as seed nodes for |
148 |
each other. What's important is that all seed nodes connections form a |
149 |
complete graph, so that the network cannot split into separate subnets |
150 |
forever. |
151 |
|
152 |
Seed nodes are represented by seed IDs. |
153 |
|
154 |
=item seed IDs - C<host:port> |
155 |
|
156 |
Seed IDs are transport endpoint(s) (usually a hostname/IP address and a |
157 |
TCP port) of nodes that should be used as seed nodes. |
158 |
|
159 |
=item global nodes |
160 |
|
161 |
An AEMP network needs a discovery service - nodes need to know how to |
162 |
connect to other nodes they only know by name. In addition, AEMP offers a |
163 |
distributed "group database", which maps group names to a list of strings |
164 |
- for example, to register worker ports. |
165 |
|
166 |
A network needs at least one global node to work, and allows every node to |
167 |
be a global node. |
168 |
|
169 |
Any node that loads the L<AnyEvent::MP::Global> module becomes a global |
170 |
node and tries to keep connections to all other nodes. So while it can |
171 |
make sense to make every node "global" in small networks, it usually makes |
172 |
sense to only make seed nodes into global nodes in large networks (nodes |
173 |
keep connections to seed nodes and global nodes, so makign them the same |
174 |
reduces overhead). |
175 |
|
176 |
=back |
177 |
|
178 |
=head1 VARIABLES/FUNCTIONS |
179 |
|
180 |
=over 4 |
181 |
|
182 |
=cut |
183 |
|
184 |
package AnyEvent::MP; |
185 |
|
186 |
use AnyEvent::MP::Config (); |
187 |
use AnyEvent::MP::Kernel; |
188 |
use AnyEvent::MP::Kernel qw( |
189 |
%NODE %PORT %PORT_DATA $UNIQ $RUNIQ $ID |
190 |
add_node load_func |
191 |
|
192 |
NODE $NODE |
193 |
configure |
194 |
node_of port_is_local |
195 |
snd kil |
196 |
db_set db_del |
197 |
db_mon db_family db_keys db_values |
198 |
); |
199 |
|
200 |
use common::sense; |
201 |
|
202 |
use Carp (); |
203 |
|
204 |
use AnyEvent (); |
205 |
use Guard (); |
206 |
|
207 |
use base "Exporter"; |
208 |
|
209 |
our $VERSION = $AnyEvent::MP::Config::VERSION; |
210 |
|
211 |
our @EXPORT = qw( |
212 |
NODE $NODE |
213 |
configure |
214 |
node_of port_is_local |
215 |
snd kil |
216 |
db_set db_del |
217 |
db_mon db_family db_keys db_values |
218 |
|
219 |
*SELF |
220 |
|
221 |
port rcv mon mon_guard psub peval spawn cal |
222 |
db_set db_del db_reg |
223 |
db_mon db_family db_keys db_values |
224 |
|
225 |
after |
226 |
); |
227 |
|
228 |
our $SELF; |
229 |
|
230 |
sub _self_die() { |
231 |
my $msg = $@; |
232 |
$msg =~ s/\n+$// unless ref $msg; |
233 |
kil $SELF, die => $msg; |
234 |
} |
235 |
|
236 |
=item $thisnode = NODE / $NODE |
237 |
|
238 |
The C<NODE> function returns, and the C<$NODE> variable contains, the node |
239 |
ID of the node running in the current process. This value is initialised by |
240 |
a call to C<configure>. |
241 |
|
242 |
=item $nodeid = node_of $port |
243 |
|
244 |
Extracts and returns the node ID from a port ID or a node ID. |
245 |
|
246 |
=item $is_local = port_is_local $port |
247 |
|
248 |
Returns true iff the port is a local port. |
249 |
|
250 |
=item configure $profile, key => value... |
251 |
|
252 |
=item configure key => value... |
253 |
|
254 |
Before a node can talk to other nodes on the network (i.e. enter |
255 |
"distributed mode") it has to configure itself - the minimum a node needs |
256 |
to know is its own name, and optionally it should know the addresses of |
257 |
some other nodes in the network to discover other nodes. |
258 |
|
259 |
This function configures a node - it must be called exactly once (or |
260 |
never) before calling other AnyEvent::MP functions. |
261 |
|
262 |
The key/value pairs are basically the same ones as documented for the |
263 |
F<aemp> command line utility (sans the set/del prefix), with these additions: |
264 |
|
265 |
=over 4 |
266 |
|
267 |
=item norc => $boolean (default false) |
268 |
|
269 |
If true, then the rc file (e.g. F<~/.perl-anyevent-mp>) will I<not> |
270 |
be consulted - all configuraiton options must be specified in the |
271 |
C<configure> call. |
272 |
|
273 |
=item force => $boolean (default false) |
274 |
|
275 |
IF true, then the values specified in the C<configure> will take |
276 |
precedence over any values configured via the rc file. The default is for |
277 |
the rc file to override any options specified in the program. |
278 |
|
279 |
=back |
280 |
|
281 |
=over 4 |
282 |
|
283 |
=item step 1, gathering configuration from profiles |
284 |
|
285 |
The function first looks up a profile in the aemp configuration (see the |
286 |
L<aemp> commandline utility). The profile name can be specified via the |
287 |
named C<profile> parameter or can simply be the first parameter). If it is |
288 |
missing, then the nodename (F<uname -n>) will be used as profile name. |
289 |
|
290 |
The profile data is then gathered as follows: |
291 |
|
292 |
First, all remaining key => value pairs (all of which are conveniently |
293 |
undocumented at the moment) will be interpreted as configuration |
294 |
data. Then they will be overwritten by any values specified in the global |
295 |
default configuration (see the F<aemp> utility), then the chain of |
296 |
profiles chosen by the profile name (and any C<parent> attributes). |
297 |
|
298 |
That means that the values specified in the profile have highest priority |
299 |
and the values specified directly via C<configure> have lowest priority, |
300 |
and can only be used to specify defaults. |
301 |
|
302 |
If the profile specifies a node ID, then this will become the node ID of |
303 |
this process. If not, then the profile name will be used as node ID, with |
304 |
a unique randoms tring (C</%u>) appended. |
305 |
|
306 |
The node ID can contain some C<%> sequences that are expanded: C<%n> |
307 |
is expanded to the local nodename, C<%u> is replaced by a random |
308 |
strign to make the node unique. For example, the F<aemp> commandline |
309 |
utility uses C<aemp/%n/%u> as nodename, which might expand to |
310 |
C<aemp/cerebro/ZQDGSIkRhEZQDGSIkRhE>. |
311 |
|
312 |
=item step 2, bind listener sockets |
313 |
|
314 |
The next step is to look up the binds in the profile, followed by binding |
315 |
aemp protocol listeners on all binds specified (it is possible and valid |
316 |
to have no binds, meaning that the node cannot be contacted form the |
317 |
outside. This means the node cannot talk to other nodes that also have no |
318 |
binds, but it can still talk to all "normal" nodes). |
319 |
|
320 |
If the profile does not specify a binds list, then a default of C<*> is |
321 |
used, meaning the node will bind on a dynamically-assigned port on every |
322 |
local IP address it finds. |
323 |
|
324 |
=item step 3, connect to seed nodes |
325 |
|
326 |
As the last step, the seed ID list from the profile is passed to the |
327 |
L<AnyEvent::MP::Global> module, which will then use it to keep |
328 |
connectivity with at least one node at any point in time. |
329 |
|
330 |
=back |
331 |
|
332 |
Example: become a distributed node using the local node name as profile. |
333 |
This should be the most common form of invocation for "daemon"-type nodes. |
334 |
|
335 |
configure |
336 |
|
337 |
Example: become a semi-anonymous node. This form is often used for |
338 |
commandline clients. |
339 |
|
340 |
configure nodeid => "myscript/%n/%u"; |
341 |
|
342 |
Example: configure a node using a profile called seed, which is suitable |
343 |
for a seed node as it binds on all local addresses on a fixed port (4040, |
344 |
customary for aemp). |
345 |
|
346 |
# use the aemp commandline utility |
347 |
# aemp profile seed binds '*:4040' |
348 |
|
349 |
# then use it |
350 |
configure profile => "seed"; |
351 |
|
352 |
# or simply use aemp from the shell again: |
353 |
# aemp run profile seed |
354 |
|
355 |
# or provide a nicer-to-remember nodeid |
356 |
# aemp run profile seed nodeid "$(hostname)" |
357 |
|
358 |
=item $SELF |
359 |
|
360 |
Contains the current port id while executing C<rcv> callbacks or C<psub> |
361 |
blocks. |
362 |
|
363 |
=item *SELF, SELF, %SELF, @SELF... |
364 |
|
365 |
Due to some quirks in how perl exports variables, it is impossible to |
366 |
just export C<$SELF>, all the symbols named C<SELF> are exported by this |
367 |
module, but only C<$SELF> is currently used. |
368 |
|
369 |
=item snd $port, type => @data |
370 |
|
371 |
=item snd $port, @msg |
372 |
|
373 |
Send the given message to the given port, which can identify either a |
374 |
local or a remote port, and must be a port ID. |
375 |
|
376 |
While the message can be almost anything, it is highly recommended to |
377 |
use a string as first element (a port ID, or some word that indicates a |
378 |
request type etc.) and to consist if only simple perl values (scalars, |
379 |
arrays, hashes) - if you think you need to pass an object, think again. |
380 |
|
381 |
The message data logically becomes read-only after a call to this |
382 |
function: modifying any argument (or values referenced by them) is |
383 |
forbidden, as there can be considerable time between the call to C<snd> |
384 |
and the time the message is actually being serialised - in fact, it might |
385 |
never be copied as within the same process it is simply handed to the |
386 |
receiving port. |
387 |
|
388 |
The type of data you can transfer depends on the transport protocol: when |
389 |
JSON is used, then only strings, numbers and arrays and hashes consisting |
390 |
of those are allowed (no objects). When Storable is used, then anything |
391 |
that Storable can serialise and deserialise is allowed, and for the local |
392 |
node, anything can be passed. Best rely only on the common denominator of |
393 |
these. |
394 |
|
395 |
=item $local_port = port |
396 |
|
397 |
Create a new local port object and returns its port ID. Initially it has |
398 |
no callbacks set and will throw an error when it receives messages. |
399 |
|
400 |
=item $local_port = port { my @msg = @_ } |
401 |
|
402 |
Creates a new local port, and returns its ID. Semantically the same as |
403 |
creating a port and calling C<rcv $port, $callback> on it. |
404 |
|
405 |
The block will be called for every message received on the port, with the |
406 |
global variable C<$SELF> set to the port ID. Runtime errors will cause the |
407 |
port to be C<kil>ed. The message will be passed as-is, no extra argument |
408 |
(i.e. no port ID) will be passed to the callback. |
409 |
|
410 |
If you want to stop/destroy the port, simply C<kil> it: |
411 |
|
412 |
my $port = port { |
413 |
my @msg = @_; |
414 |
... |
415 |
kil $SELF; |
416 |
}; |
417 |
|
418 |
=cut |
419 |
|
420 |
sub rcv($@); |
421 |
|
422 |
my $KILME = sub { |
423 |
(my $tag = substr $_[0], 0, 30) =~ s/([\x20-\x7e])/./g; |
424 |
kil $SELF, unhandled_message => "no callback found for message '$tag'"; |
425 |
}; |
426 |
|
427 |
sub port(;&) { |
428 |
my $id = $UNIQ . ++$ID; |
429 |
my $port = "$NODE#$id"; |
430 |
|
431 |
rcv $port, shift || $KILME; |
432 |
|
433 |
$port |
434 |
} |
435 |
|
436 |
=item rcv $local_port, $callback->(@msg) |
437 |
|
438 |
Replaces the default callback on the specified port. There is no way to |
439 |
remove the default callback: use C<sub { }> to disable it, or better |
440 |
C<kil> the port when it is no longer needed. |
441 |
|
442 |
The global C<$SELF> (exported by this module) contains C<$port> while |
443 |
executing the callback. Runtime errors during callback execution will |
444 |
result in the port being C<kil>ed. |
445 |
|
446 |
The default callback receives all messages not matched by a more specific |
447 |
C<tag> match. |
448 |
|
449 |
=item rcv $local_port, tag => $callback->(@msg_without_tag), ... |
450 |
|
451 |
Register (or replace) callbacks to be called on messages starting with the |
452 |
given tag on the given port (and return the port), or unregister it (when |
453 |
C<$callback> is C<$undef> or missing). There can only be one callback |
454 |
registered for each tag. |
455 |
|
456 |
The original message will be passed to the callback, after the first |
457 |
element (the tag) has been removed. The callback will use the same |
458 |
environment as the default callback (see above). |
459 |
|
460 |
Example: create a port and bind receivers on it in one go. |
461 |
|
462 |
my $port = rcv port, |
463 |
msg1 => sub { ... }, |
464 |
msg2 => sub { ... }, |
465 |
; |
466 |
|
467 |
Example: create a port, bind receivers and send it in a message elsewhere |
468 |
in one go: |
469 |
|
470 |
snd $otherport, reply => |
471 |
rcv port, |
472 |
msg1 => sub { ... }, |
473 |
... |
474 |
; |
475 |
|
476 |
Example: temporarily register a rcv callback for a tag matching some port |
477 |
(e.g. for an rpc reply) and unregister it after a message was received. |
478 |
|
479 |
rcv $port, $otherport => sub { |
480 |
my @reply = @_; |
481 |
|
482 |
rcv $SELF, $otherport; |
483 |
}; |
484 |
|
485 |
=cut |
486 |
|
487 |
sub rcv($@) { |
488 |
my $port = shift; |
489 |
my ($nodeid, $portid) = split /#/, $port, 2; |
490 |
|
491 |
$nodeid eq $NODE |
492 |
or Carp::croak "$port: rcv can only be called on local ports, caught"; |
493 |
|
494 |
while (@_) { |
495 |
if (ref $_[0]) { |
496 |
if (my $self = $PORT_DATA{$portid}) { |
497 |
"AnyEvent::MP::Port" eq ref $self |
498 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
499 |
|
500 |
$self->[0] = shift; |
501 |
} else { |
502 |
my $cb = shift; |
503 |
$PORT{$portid} = sub { |
504 |
local $SELF = $port; |
505 |
eval { &$cb }; _self_die if $@; |
506 |
}; |
507 |
} |
508 |
} elsif (defined $_[0]) { |
509 |
my $self = $PORT_DATA{$portid} ||= do { |
510 |
my $self = bless [$PORT{$portid} || sub { }, { }, $port], "AnyEvent::MP::Port"; |
511 |
|
512 |
$PORT{$portid} = sub { |
513 |
local $SELF = $port; |
514 |
|
515 |
if (my $cb = $self->[1]{$_[0]}) { |
516 |
shift; |
517 |
eval { &$cb }; _self_die if $@; |
518 |
} else { |
519 |
&{ $self->[0] }; |
520 |
} |
521 |
}; |
522 |
|
523 |
$self |
524 |
}; |
525 |
|
526 |
"AnyEvent::MP::Port" eq ref $self |
527 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
528 |
|
529 |
my ($tag, $cb) = splice @_, 0, 2; |
530 |
|
531 |
if (defined $cb) { |
532 |
$self->[1]{$tag} = $cb; |
533 |
} else { |
534 |
delete $self->[1]{$tag}; |
535 |
} |
536 |
} |
537 |
} |
538 |
|
539 |
$port |
540 |
} |
541 |
|
542 |
=item peval $port, $coderef[, @args] |
543 |
|
544 |
Evaluates the given C<$codref> within the contetx of C<$port>, that is, |
545 |
when the code throews an exception the C<$port> will be killed. |
546 |
|
547 |
Any remaining args will be passed to the callback. Any return values will |
548 |
be returned to the caller. |
549 |
|
550 |
This is useful when you temporarily want to execute code in the context of |
551 |
a port. |
552 |
|
553 |
Example: create a port and run some initialisation code in it's context. |
554 |
|
555 |
my $port = port { ... }; |
556 |
|
557 |
peval $port, sub { |
558 |
init |
559 |
or die "unable to init"; |
560 |
}; |
561 |
|
562 |
=cut |
563 |
|
564 |
sub peval($$) { |
565 |
local $SELF = shift; |
566 |
my $cb = shift; |
567 |
|
568 |
if (wantarray) { |
569 |
my @res = eval { &$cb }; |
570 |
_self_die if $@; |
571 |
@res |
572 |
} else { |
573 |
my $res = eval { &$cb }; |
574 |
_self_die if $@; |
575 |
$res |
576 |
} |
577 |
} |
578 |
|
579 |
=item $closure = psub { BLOCK } |
580 |
|
581 |
Remembers C<$SELF> and creates a closure out of the BLOCK. When the |
582 |
closure is executed, sets up the environment in the same way as in C<rcv> |
583 |
callbacks, i.e. runtime errors will cause the port to get C<kil>ed. |
584 |
|
585 |
The effect is basically as if it returned C<< sub { peval $SELF, sub { |
586 |
BLOCK }, @_ } >>. |
587 |
|
588 |
This is useful when you register callbacks from C<rcv> callbacks: |
589 |
|
590 |
rcv delayed_reply => sub { |
591 |
my ($delay, @reply) = @_; |
592 |
my $timer = AE::timer $delay, 0, psub { |
593 |
snd @reply, $SELF; |
594 |
}; |
595 |
}; |
596 |
|
597 |
=cut |
598 |
|
599 |
sub psub(&) { |
600 |
my $cb = shift; |
601 |
|
602 |
my $port = $SELF |
603 |
or Carp::croak "psub can only be called from within rcv or psub callbacks, not"; |
604 |
|
605 |
sub { |
606 |
local $SELF = $port; |
607 |
|
608 |
if (wantarray) { |
609 |
my @res = eval { &$cb }; |
610 |
_self_die if $@; |
611 |
@res |
612 |
} else { |
613 |
my $res = eval { &$cb }; |
614 |
_self_die if $@; |
615 |
$res |
616 |
} |
617 |
} |
618 |
} |
619 |
|
620 |
=item $guard = mon $port, $rcvport # kill $rcvport when $port dies |
621 |
|
622 |
=item $guard = mon $port # kill $SELF when $port dies |
623 |
|
624 |
=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies |
625 |
|
626 |
=item $guard = mon $port, $rcvport, @msg # send a message when $port dies |
627 |
|
628 |
Monitor the given port and do something when the port is killed or |
629 |
messages to it were lost, and optionally return a guard that can be used |
630 |
to stop monitoring again. |
631 |
|
632 |
The first two forms distinguish between "normal" and "abnormal" kil's: |
633 |
|
634 |
In the first form (another port given), if the C<$port> is C<kil>'ed with |
635 |
a non-empty reason, the other port (C<$rcvport>) will be kil'ed with the |
636 |
same reason. That is, on "normal" kil's nothing happens, while under all |
637 |
other conditions, the other port is killed with the same reason. |
638 |
|
639 |
The second form (kill self) is the same as the first form, except that |
640 |
C<$rvport> defaults to C<$SELF>. |
641 |
|
642 |
The remaining forms don't distinguish between "normal" and "abnormal" kil's |
643 |
- it's up to the callback or receiver to check whether the C<@reason> is |
644 |
empty and act accordingly. |
645 |
|
646 |
In the third form (callback), the callback is simply called with any |
647 |
number of C<@reason> elements (empty @reason means that the port was deleted |
648 |
"normally"). Note also that I<< the callback B<must> never die >>, so use |
649 |
C<eval> if unsure. |
650 |
|
651 |
In the last form (message), a message of the form C<$rcvport, @msg, |
652 |
@reason> will be C<snd>. |
653 |
|
654 |
Monitoring-actions are one-shot: once messages are lost (and a monitoring |
655 |
alert was raised), they are removed and will not trigger again, even if it |
656 |
turns out that the port is still alive. |
657 |
|
658 |
As a rule of thumb, monitoring requests should always monitor a remote |
659 |
port locally (using a local C<$rcvport> or a callback). The reason is that |
660 |
kill messages might get lost, just like any other message. Another less |
661 |
obvious reason is that even monitoring requests can get lost (for example, |
662 |
when the connection to the other node goes down permanently). When |
663 |
monitoring a port locally these problems do not exist. |
664 |
|
665 |
C<mon> effectively guarantees that, in the absence of hardware failures, |
666 |
after starting the monitor, either all messages sent to the port will |
667 |
arrive, or the monitoring action will be invoked after possible message |
668 |
loss has been detected. No messages will be lost "in between" (after |
669 |
the first lost message no further messages will be received by the |
670 |
port). After the monitoring action was invoked, further messages might get |
671 |
delivered again. |
672 |
|
673 |
Inter-host-connection timeouts and monitoring depend on the transport |
674 |
used. The only transport currently implemented is TCP, and AnyEvent::MP |
675 |
relies on TCP to detect node-downs (this can take 10-15 minutes on a |
676 |
non-idle connection, and usually around two hours for idle connections). |
677 |
|
678 |
This means that monitoring is good for program errors and cleaning up |
679 |
stuff eventually, but they are no replacement for a timeout when you need |
680 |
to ensure some maximum latency. |
681 |
|
682 |
Example: call a given callback when C<$port> is killed. |
683 |
|
684 |
mon $port, sub { warn "port died because of <@_>\n" }; |
685 |
|
686 |
Example: kill ourselves when C<$port> is killed abnormally. |
687 |
|
688 |
mon $port; |
689 |
|
690 |
Example: send us a restart message when another C<$port> is killed. |
691 |
|
692 |
mon $port, $self => "restart"; |
693 |
|
694 |
=cut |
695 |
|
696 |
sub mon { |
697 |
my ($nodeid, $port) = split /#/, shift, 2; |
698 |
|
699 |
my $node = $NODE{$nodeid} || add_node $nodeid; |
700 |
|
701 |
my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,'; |
702 |
|
703 |
unless (ref $cb) { |
704 |
if (@_) { |
705 |
# send a kill info message |
706 |
my (@msg) = ($cb, @_); |
707 |
$cb = sub { snd @msg, @_ }; |
708 |
} else { |
709 |
# simply kill other port |
710 |
my $port = $cb; |
711 |
$cb = sub { kil $port, @_ if @_ }; |
712 |
} |
713 |
} |
714 |
|
715 |
$node->monitor ($port, $cb); |
716 |
|
717 |
defined wantarray |
718 |
and ($cb += 0, Guard::guard { $node->unmonitor ($port, $cb) }) |
719 |
} |
720 |
|
721 |
=item $guard = mon_guard $port, $ref, $ref... |
722 |
|
723 |
Monitors the given C<$port> and keeps the passed references. When the port |
724 |
is killed, the references will be freed. |
725 |
|
726 |
Optionally returns a guard that will stop the monitoring. |
727 |
|
728 |
This function is useful when you create e.g. timers or other watchers and |
729 |
want to free them when the port gets killed (note the use of C<psub>): |
730 |
|
731 |
$port->rcv (start => sub { |
732 |
my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub { |
733 |
undef $timer if 0.9 < rand; |
734 |
}); |
735 |
}); |
736 |
|
737 |
=cut |
738 |
|
739 |
sub mon_guard { |
740 |
my ($port, @refs) = @_; |
741 |
|
742 |
#TODO: mon-less form? |
743 |
|
744 |
mon $port, sub { 0 && @refs } |
745 |
} |
746 |
|
747 |
=item kil $port[, @reason] |
748 |
|
749 |
Kill the specified port with the given C<@reason>. |
750 |
|
751 |
If no C<@reason> is specified, then the port is killed "normally" - |
752 |
monitor callback will be invoked, but the kil will not cause linked ports |
753 |
(C<mon $mport, $lport> form) to get killed. |
754 |
|
755 |
If a C<@reason> is specified, then linked ports (C<mon $mport, $lport> |
756 |
form) get killed with the same reason. |
757 |
|
758 |
Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks |
759 |
will be reported as reason C<< die => $@ >>. |
760 |
|
761 |
Transport/communication errors are reported as C<< transport_error => |
762 |
$message >>. |
763 |
|
764 |
Common idioms: |
765 |
|
766 |
# silently remove yourself, do not kill linked ports |
767 |
kil $SELF; |
768 |
|
769 |
# report a failure in some detail |
770 |
kil $SELF, failure_mode_1 => "it failed with too high temperature"; |
771 |
|
772 |
# do not waste much time with killing, just die when something goes wrong |
773 |
open my $fh, "<file" |
774 |
or die "file: $!"; |
775 |
|
776 |
=item $port = spawn $node, $initfunc[, @initdata] |
777 |
|
778 |
Creates a port on the node C<$node> (which can also be a port ID, in which |
779 |
case it's the node where that port resides). |
780 |
|
781 |
The port ID of the newly created port is returned immediately, and it is |
782 |
possible to immediately start sending messages or to monitor the port. |
783 |
|
784 |
After the port has been created, the init function is called on the remote |
785 |
node, in the same context as a C<rcv> callback. This function must be a |
786 |
fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To |
787 |
specify a function in the main program, use C<::name>. |
788 |
|
789 |
If the function doesn't exist, then the node tries to C<require> |
790 |
the package, then the package above the package and so on (e.g. |
791 |
C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
792 |
exists or it runs out of package names. |
793 |
|
794 |
The init function is then called with the newly-created port as context |
795 |
object (C<$SELF>) and the C<@initdata> values as arguments. It I<must> |
796 |
call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise |
797 |
the port might not get created. |
798 |
|
799 |
A common idiom is to pass a local port, immediately monitor the spawned |
800 |
port, and in the remote init function, immediately monitor the passed |
801 |
local port. This two-way monitoring ensures that both ports get cleaned up |
802 |
when there is a problem. |
803 |
|
804 |
C<spawn> guarantees that the C<$initfunc> has no visible effects on the |
805 |
caller before C<spawn> returns (by delaying invocation when spawn is |
806 |
called for the local node). |
807 |
|
808 |
Example: spawn a chat server port on C<$othernode>. |
809 |
|
810 |
# this node, executed from within a port context: |
811 |
my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
812 |
mon $server; |
813 |
|
814 |
# init function on C<$othernode> |
815 |
sub connect { |
816 |
my ($srcport) = @_; |
817 |
|
818 |
mon $srcport; |
819 |
|
820 |
rcv $SELF, sub { |
821 |
... |
822 |
}; |
823 |
} |
824 |
|
825 |
=cut |
826 |
|
827 |
sub _spawn { |
828 |
my $port = shift; |
829 |
my $init = shift; |
830 |
|
831 |
# rcv will create the actual port |
832 |
local $SELF = "$NODE#$port"; |
833 |
eval { |
834 |
&{ load_func $init } |
835 |
}; |
836 |
_self_die if $@; |
837 |
} |
838 |
|
839 |
sub spawn(@) { |
840 |
my ($nodeid, undef) = split /#/, shift, 2; |
841 |
|
842 |
my $id = $RUNIQ . ++$ID; |
843 |
|
844 |
$_[0] =~ /::/ |
845 |
or Carp::croak "spawn init function must be a fully-qualified name, caught"; |
846 |
|
847 |
snd_to_func $nodeid, "AnyEvent::MP::_spawn" => $id, @_; |
848 |
|
849 |
"$nodeid#$id" |
850 |
} |
851 |
|
852 |
|
853 |
=item after $timeout, @msg |
854 |
|
855 |
=item after $timeout, $callback |
856 |
|
857 |
Either sends the given message, or call the given callback, after the |
858 |
specified number of seconds. |
859 |
|
860 |
This is simply a utility function that comes in handy at times - the |
861 |
AnyEvent::MP author is not convinced of the wisdom of having it, though, |
862 |
so it may go away in the future. |
863 |
|
864 |
=cut |
865 |
|
866 |
sub after($@) { |
867 |
my ($timeout, @action) = @_; |
868 |
|
869 |
my $t; $t = AE::timer $timeout, 0, sub { |
870 |
undef $t; |
871 |
ref $action[0] |
872 |
? $action[0]() |
873 |
: snd @action; |
874 |
}; |
875 |
} |
876 |
|
877 |
#=item $cb2 = timeout $seconds, $cb[, @args] |
878 |
|
879 |
=item cal $port, @msg, $callback[, $timeout] |
880 |
|
881 |
A simple form of RPC - sends a message to the given C<$port> with the |
882 |
given contents (C<@msg>), but adds a reply port to the message. |
883 |
|
884 |
The reply port is created temporarily just for the purpose of receiving |
885 |
the reply, and will be C<kil>ed when no longer needed. |
886 |
|
887 |
A reply message sent to the port is passed to the C<$callback> as-is. |
888 |
|
889 |
If an optional time-out (in seconds) is given and it is not C<undef>, |
890 |
then the callback will be called without any arguments after the time-out |
891 |
elapsed and the port is C<kil>ed. |
892 |
|
893 |
If no time-out is given (or it is C<undef>), then the local port will |
894 |
monitor the remote port instead, so it eventually gets cleaned-up. |
895 |
|
896 |
Currently this function returns the temporary port, but this "feature" |
897 |
might go in future versions unless you can make a convincing case that |
898 |
this is indeed useful for something. |
899 |
|
900 |
=cut |
901 |
|
902 |
sub cal(@) { |
903 |
my $timeout = ref $_[-1] ? undef : pop; |
904 |
my $cb = pop; |
905 |
|
906 |
my $port = port { |
907 |
undef $timeout; |
908 |
kil $SELF; |
909 |
&$cb; |
910 |
}; |
911 |
|
912 |
if (defined $timeout) { |
913 |
$timeout = AE::timer $timeout, 0, sub { |
914 |
undef $timeout; |
915 |
kil $port; |
916 |
$cb->(); |
917 |
}; |
918 |
} else { |
919 |
mon $_[0], sub { |
920 |
kil $port; |
921 |
$cb->(); |
922 |
}; |
923 |
} |
924 |
|
925 |
push @_, $port; |
926 |
&snd; |
927 |
|
928 |
$port |
929 |
} |
930 |
|
931 |
=back |
932 |
|
933 |
=head1 DISTRIBUTED DATABASE |
934 |
|
935 |
AnyEvent::MP comes with a simple distributed database. The database will |
936 |
be mirrored asynchronously on all global nodes. Other nodes bind to one |
937 |
of the global nodes for their needs. Every node has a "local database" |
938 |
which contains all the values that are set locally. All local databases |
939 |
are merged together to form the global database, which can be queried. |
940 |
|
941 |
The database structure is that of a two-level hash - the database hash |
942 |
contains hashes which contain values, similarly to a perl hash of hashes, |
943 |
i.e.: |
944 |
|
945 |
$DATABASE{$family}{$subkey} = $value |
946 |
|
947 |
The top level hash key is called "family", and the second-level hash key |
948 |
is called "subkey" or simply "key". |
949 |
|
950 |
The family must be alphanumeric, i.e. start with a letter and consist |
951 |
of letters, digits, underscores and colons (C<[A-Za-z][A-Za-z0-9_:]*>, |
952 |
pretty much like Perl module names. |
953 |
|
954 |
As the family namespace is global, it is recommended to prefix family names |
955 |
with the name of the application or module using it. |
956 |
|
957 |
The subkeys must be non-empty strings, with no further restrictions. |
958 |
|
959 |
The values should preferably be strings, but other perl scalars should |
960 |
work as well (such as C<undef>, arrays and hashes). |
961 |
|
962 |
Every database entry is owned by one node - adding the same family/subkey |
963 |
combination on multiple nodes will not cause discomfort for AnyEvent::MP, |
964 |
but the result might be nondeterministic, i.e. the key might have |
965 |
different values on different nodes. |
966 |
|
967 |
Different subkeys in the same family can be owned by different nodes |
968 |
without problems, and in fact, this is the common method to create worker |
969 |
pools. For example, a worker port for image scaling might do this: |
970 |
|
971 |
db_set my_image_scalers => $port; |
972 |
|
973 |
And clients looking for an image scaler will want to get the |
974 |
C<my_image_scalers> keys from time to time: |
975 |
|
976 |
db_keys my_image_scalers => sub { |
977 |
@ports = @{ $_[0] }; |
978 |
}; |
979 |
|
980 |
Or better yet, they want to monitor the database family, so they always |
981 |
have a reasonable up-to-date copy: |
982 |
|
983 |
db_mon my_image_scalers => sub { |
984 |
@ports = keys %{ $_[0] }; |
985 |
}; |
986 |
|
987 |
In general, you can set or delete single subkeys, but query and monitor |
988 |
whole families only. |
989 |
|
990 |
If you feel the need to monitor or query a single subkey, try giving it |
991 |
it's own family. |
992 |
|
993 |
=over |
994 |
|
995 |
=item $guard = db_set $family => $subkey [=> $value] |
996 |
|
997 |
Sets (or replaces) a key to the database - if C<$value> is omitted, |
998 |
C<undef> is used instead. |
999 |
|
1000 |
When called in non-void context, C<db_set> returns a guard that |
1001 |
automatically calls C<db_del> when it is destroyed. |
1002 |
|
1003 |
=item db_del $family => $subkey... |
1004 |
|
1005 |
Deletes one or more subkeys from the database family. |
1006 |
|
1007 |
=item $guard = db_reg $family => $port => $value |
1008 |
|
1009 |
=item $guard = db_reg $family => $port |
1010 |
|
1011 |
=item $guard = db_reg $family |
1012 |
|
1013 |
Registers a port in the given family and optionally returns a guard to |
1014 |
remove it. |
1015 |
|
1016 |
This function basically does the same as: |
1017 |
|
1018 |
db_set $family => $port => $value |
1019 |
|
1020 |
Except that the port is monitored and automatically removed from the |
1021 |
database family when it is kil'ed. |
1022 |
|
1023 |
If C<$value> is missing, C<undef> is used. If C<$port> is missing, then |
1024 |
C<$SELF> is used. |
1025 |
|
1026 |
This function is most useful to register a port in some port group (which |
1027 |
is just another name for a database family), and have it removed when the |
1028 |
port is gone. This works best when the port is a local port. |
1029 |
|
1030 |
=cut |
1031 |
|
1032 |
sub db_reg($$;$) { |
1033 |
my $family = shift; |
1034 |
my $port = @_ ? shift : $SELF; |
1035 |
|
1036 |
my $clr = sub { db_del $family => $port }; |
1037 |
mon $port, $clr; |
1038 |
|
1039 |
db_set $family => $port => $_[0]; |
1040 |
|
1041 |
defined wantarray |
1042 |
and &Guard::guard ($clr) |
1043 |
} |
1044 |
|
1045 |
=item db_family $family => $cb->(\%familyhash) |
1046 |
|
1047 |
Queries the named database C<$family> and call the callback with the |
1048 |
family represented as a hash. You can keep and freely modify the hash. |
1049 |
|
1050 |
=item db_keys $family => $cb->(\@keys) |
1051 |
|
1052 |
Same as C<db_family>, except it only queries the family I<subkeys> and passes |
1053 |
them as array reference to the callback. |
1054 |
|
1055 |
=item db_values $family => $cb->(\@values) |
1056 |
|
1057 |
Same as C<db_family>, except it only queries the family I<values> and passes them |
1058 |
as array reference to the callback. |
1059 |
|
1060 |
=item $guard = db_mon $family => $cb->(\%familyhash, \@added, \@changed, \@deleted) |
1061 |
|
1062 |
Creates a monitor on the given database family. Each time a key is set |
1063 |
or or is deleted the callback is called with a hash containing the |
1064 |
database family and three lists of added, changed and deleted subkeys, |
1065 |
respectively. If no keys have changed then the array reference might be |
1066 |
C<undef> or even missing. |
1067 |
|
1068 |
If not called in void context, a guard object is returned that, when |
1069 |
destroyed, stops the monitor. |
1070 |
|
1071 |
The family hash reference and the key arrays belong to AnyEvent::MP and |
1072 |
B<must not be modified or stored> by the callback. When in doubt, make a |
1073 |
copy. |
1074 |
|
1075 |
As soon as possible after the monitoring starts, the callback will be |
1076 |
called with the intiial contents of the family, even if it is empty, |
1077 |
i.e. there will always be a timely call to the callback with the current |
1078 |
contents. |
1079 |
|
1080 |
It is possible that the callback is called with a change event even though |
1081 |
the subkey is already present and the value has not changed. |
1082 |
|
1083 |
The monitoring stops when the guard object is destroyed. |
1084 |
|
1085 |
Example: on every change to the family "mygroup", print out all keys. |
1086 |
|
1087 |
my $guard = db_mon mygroup => sub { |
1088 |
my ($family, $a, $c, $d) = @_; |
1089 |
print "mygroup members: ", (join " ", keys %$family), "\n"; |
1090 |
}; |
1091 |
|
1092 |
Exmaple: wait until the family "My::Module::workers" is non-empty. |
1093 |
|
1094 |
my $guard; $guard = db_mon My::Module::workers => sub { |
1095 |
my ($family, $a, $c, $d) = @_; |
1096 |
return unless %$family; |
1097 |
undef $guard; |
1098 |
print "My::Module::workers now nonempty\n"; |
1099 |
}; |
1100 |
|
1101 |
Example: print all changes to the family "AnyRvent::Fantasy::Module". |
1102 |
|
1103 |
my $guard = db_mon AnyRvent::Fantasy::Module => sub { |
1104 |
my ($family, $a, $c, $d) = @_; |
1105 |
|
1106 |
print "+$_=$family->{$_}\n" for @$a; |
1107 |
print "*$_=$family->{$_}\n" for @$c; |
1108 |
print "-$_=$family->{$_}\n" for @$d; |
1109 |
}; |
1110 |
|
1111 |
=cut |
1112 |
|
1113 |
=back |
1114 |
|
1115 |
=head1 AnyEvent::MP vs. Distributed Erlang |
1116 |
|
1117 |
AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
1118 |
== aemp node, Erlang process == aemp port), so many of the documents and |
1119 |
programming techniques employed by Erlang apply to AnyEvent::MP. Here is a |
1120 |
sample: |
1121 |
|
1122 |
http://www.erlang.se/doc/programming_rules.shtml |
1123 |
http://erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4 |
1124 |
http://erlang.org/download/erlang-book-part1.pdf # chapters 5 and 6 |
1125 |
http://erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5 |
1126 |
|
1127 |
Despite the similarities, there are also some important differences: |
1128 |
|
1129 |
=over 4 |
1130 |
|
1131 |
=item * Node IDs are arbitrary strings in AEMP. |
1132 |
|
1133 |
Erlang relies on special naming and DNS to work everywhere in the same |
1134 |
way. AEMP relies on each node somehow knowing its own address(es) (e.g. by |
1135 |
configuration or DNS), and possibly the addresses of some seed nodes, but |
1136 |
will otherwise discover other nodes (and their IDs) itself. |
1137 |
|
1138 |
=item * Erlang has a "remote ports are like local ports" philosophy, AEMP |
1139 |
uses "local ports are like remote ports". |
1140 |
|
1141 |
The failure modes for local ports are quite different (runtime errors |
1142 |
only) then for remote ports - when a local port dies, you I<know> it dies, |
1143 |
when a connection to another node dies, you know nothing about the other |
1144 |
port. |
1145 |
|
1146 |
Erlang pretends remote ports are as reliable as local ports, even when |
1147 |
they are not. |
1148 |
|
1149 |
AEMP encourages a "treat remote ports differently" philosophy, with local |
1150 |
ports being the special case/exception, where transport errors cannot |
1151 |
occur. |
1152 |
|
1153 |
=item * Erlang uses processes and a mailbox, AEMP does not queue. |
1154 |
|
1155 |
Erlang uses processes that selectively receive messages out of order, and |
1156 |
therefore needs a queue. AEMP is event based, queuing messages would serve |
1157 |
no useful purpose. For the same reason the pattern-matching abilities |
1158 |
of AnyEvent::MP are more limited, as there is little need to be able to |
1159 |
filter messages without dequeuing them. |
1160 |
|
1161 |
This is not a philosophical difference, but simply stems from AnyEvent::MP |
1162 |
being event-based, while Erlang is process-based. |
1163 |
|
1164 |
You cna have a look at L<Coro::MP> for a more Erlang-like process model on |
1165 |
top of AEMP and Coro threads. |
1166 |
|
1167 |
=item * Erlang sends are synchronous, AEMP sends are asynchronous. |
1168 |
|
1169 |
Sending messages in Erlang is synchronous and blocks the process until |
1170 |
a conenction has been established and the message sent (and so does not |
1171 |
need a queue that can overflow). AEMP sends return immediately, connection |
1172 |
establishment is handled in the background. |
1173 |
|
1174 |
=item * Erlang suffers from silent message loss, AEMP does not. |
1175 |
|
1176 |
Erlang implements few guarantees on messages delivery - messages can get |
1177 |
lost without any of the processes realising it (i.e. you send messages a, |
1178 |
b, and c, and the other side only receives messages a and c). |
1179 |
|
1180 |
AEMP guarantees (modulo hardware errors) correct ordering, and the |
1181 |
guarantee that after one message is lost, all following ones sent to the |
1182 |
same port are lost as well, until monitoring raises an error, so there are |
1183 |
no silent "holes" in the message sequence. |
1184 |
|
1185 |
If you want your software to be very reliable, you have to cope with |
1186 |
corrupted and even out-of-order messages in both Erlang and AEMP. AEMP |
1187 |
simply tries to work better in common error cases, such as when a network |
1188 |
link goes down. |
1189 |
|
1190 |
=item * Erlang can send messages to the wrong port, AEMP does not. |
1191 |
|
1192 |
In Erlang it is quite likely that a node that restarts reuses an Erlang |
1193 |
process ID known to other nodes for a completely different process, |
1194 |
causing messages destined for that process to end up in an unrelated |
1195 |
process. |
1196 |
|
1197 |
AEMP does not reuse port IDs, so old messages or old port IDs floating |
1198 |
around in the network will not be sent to an unrelated port. |
1199 |
|
1200 |
=item * Erlang uses unprotected connections, AEMP uses secure |
1201 |
authentication and can use TLS. |
1202 |
|
1203 |
AEMP can use a proven protocol - TLS - to protect connections and |
1204 |
securely authenticate nodes. |
1205 |
|
1206 |
=item * The AEMP protocol is optimised for both text-based and binary |
1207 |
communications. |
1208 |
|
1209 |
The AEMP protocol, unlike the Erlang protocol, supports both programming |
1210 |
language independent text-only protocols (good for debugging), and binary, |
1211 |
language-specific serialisers (e.g. Storable). By default, unless TLS is |
1212 |
used, the protocol is actually completely text-based. |
1213 |
|
1214 |
It has also been carefully designed to be implementable in other languages |
1215 |
with a minimum of work while gracefully degrading functionality to make the |
1216 |
protocol simple. |
1217 |
|
1218 |
=item * AEMP has more flexible monitoring options than Erlang. |
1219 |
|
1220 |
In Erlang, you can chose to receive I<all> exit signals as messages or |
1221 |
I<none>, there is no in-between, so monitoring single Erlang processes is |
1222 |
difficult to implement. |
1223 |
|
1224 |
Monitoring in AEMP is more flexible than in Erlang, as one can choose |
1225 |
between automatic kill, exit message or callback on a per-port basis. |
1226 |
|
1227 |
=item * Erlang tries to hide remote/local connections, AEMP does not. |
1228 |
|
1229 |
Monitoring in Erlang is not an indicator of process death/crashes, in the |
1230 |
same way as linking is (except linking is unreliable in Erlang). |
1231 |
|
1232 |
In AEMP, you don't "look up" registered port names or send to named ports |
1233 |
that might or might not be persistent. Instead, you normally spawn a port |
1234 |
on the remote node. The init function monitors you, and you monitor the |
1235 |
remote port. Since both monitors are local to the node, they are much more |
1236 |
reliable (no need for C<spawn_link>). |
1237 |
|
1238 |
This also saves round-trips and avoids sending messages to the wrong port |
1239 |
(hard to do in Erlang). |
1240 |
|
1241 |
=back |
1242 |
|
1243 |
=head1 RATIONALE |
1244 |
|
1245 |
=over 4 |
1246 |
|
1247 |
=item Why strings for port and node IDs, why not objects? |
1248 |
|
1249 |
We considered "objects", but found that the actual number of methods |
1250 |
that can be called are quite low. Since port and node IDs travel over |
1251 |
the network frequently, the serialising/deserialising would add lots of |
1252 |
overhead, as well as having to keep a proxy object everywhere. |
1253 |
|
1254 |
Strings can easily be printed, easily serialised etc. and need no special |
1255 |
procedures to be "valid". |
1256 |
|
1257 |
And as a result, a port with just a default receiver consists of a single |
1258 |
code reference stored in a global hash - it can't become much cheaper. |
1259 |
|
1260 |
=item Why favour JSON, why not a real serialising format such as Storable? |
1261 |
|
1262 |
In fact, any AnyEvent::MP node will happily accept Storable as framing |
1263 |
format, but currently there is no way to make a node use Storable by |
1264 |
default (although all nodes will accept it). |
1265 |
|
1266 |
The default framing protocol is JSON because a) JSON::XS is many times |
1267 |
faster for small messages and b) most importantly, after years of |
1268 |
experience we found that object serialisation is causing more problems |
1269 |
than it solves: Just like function calls, objects simply do not travel |
1270 |
easily over the network, mostly because they will always be a copy, so you |
1271 |
always have to re-think your design. |
1272 |
|
1273 |
Keeping your messages simple, concentrating on data structures rather than |
1274 |
objects, will keep your messages clean, tidy and efficient. |
1275 |
|
1276 |
=back |
1277 |
|
1278 |
=head1 PORTING FROM AnyEvent::MP VERSION 1.X |
1279 |
|
1280 |
AEMP version 2 has a few major incompatible changes compared to version 1: |
1281 |
|
1282 |
=over 4 |
1283 |
|
1284 |
=item AnyEvent::MP::Global no longer has group management functions. |
1285 |
|
1286 |
At least not officially - the grp_* functions are still exported and might |
1287 |
work, but they will be removed in some later release. |
1288 |
|
1289 |
AnyEvent::MP now comes with a distributed database that is more |
1290 |
powerful. Its database families map closely to port groups, but the API |
1291 |
has changed (the functions are also now exported by AnyEvent::MP). Here is |
1292 |
a rough porting guide: |
1293 |
|
1294 |
grp_reg $group, $port # old |
1295 |
db_reg $group, $port # new |
1296 |
|
1297 |
$list = grp_get $group # old |
1298 |
db_keys $group, sub { my $list = shift } # new |
1299 |
|
1300 |
grp_mon $group, $cb->(\@ports, $add, $del) # old |
1301 |
db_mon $group, $cb->(\%ports, $add, $change, $del) # new |
1302 |
|
1303 |
C<grp_reg> is a no-brainer (just replace by C<db_reg>), but C<grp_get> is |
1304 |
no longer instant, because the local node might not have a copy of the |
1305 |
group. You can either modify your code to allow for a callback, or use |
1306 |
C<db_mon> to keep an updated copy of the group: |
1307 |
|
1308 |
my $local_group_copy; |
1309 |
db_mon $group => sub { $local_group_copy = $_[0] }; |
1310 |
|
1311 |
# now "keys %$local_group_copy" always returns the most up-to-date |
1312 |
# list of ports in the group. |
1313 |
|
1314 |
C<grp_mon> can be replaced by C<db_mon> with minor changes - C<db_mon> |
1315 |
passes a hash as first argument, and an extra C<$chg> argument that can be |
1316 |
ignored: |
1317 |
|
1318 |
db_mon $group => sub { |
1319 |
my ($ports, $add, $chg, $lde) = @_; |
1320 |
$ports = [keys %$ports]; |
1321 |
|
1322 |
# now $ports, $add and $del are the same as |
1323 |
# were originally passed by grp_mon. |
1324 |
... |
1325 |
}; |
1326 |
|
1327 |
=item Nodes not longer connect to all other nodes. |
1328 |
|
1329 |
In AEMP 1.x, every node automatically loads the L<AnyEvent::MP::Global> |
1330 |
module, which in turn would create connections to all other nodes in the |
1331 |
network (helped by the seed nodes). |
1332 |
|
1333 |
In version 2.x, global nodes still connect to all other global nodes, but |
1334 |
other nodes don't - now every node either is a global node itself, or |
1335 |
attaches itself to another global node. |
1336 |
|
1337 |
If a node isn't a global node itself, then it attaches itself to one |
1338 |
of its seed nodes. If that seed node isn't a global node yet, it will |
1339 |
automatically be upgraded to a global node. |
1340 |
|
1341 |
So in many cases, nothing needs to be changed - one just has to make sure |
1342 |
that all seed nodes are meshed together with the other seed nodes (as with |
1343 |
AEMP 1.x), and other nodes specify them as seed nodes. This is most easily |
1344 |
achieved by specifying the same set of seed nodes for all nodes in the |
1345 |
network. |
1346 |
|
1347 |
Not opening a connection to every other node is usually an advantage, |
1348 |
except when you need the lower latency of an already established |
1349 |
connection. To ensure a node establishes a connection to another node, |
1350 |
you can monitor the node port (C<mon $node, ...>), which will attempt to |
1351 |
create the connection (and notify you when the connection fails). |
1352 |
|
1353 |
=item Listener-less nodes (nodes without binds) are gone. |
1354 |
|
1355 |
And are not coming back, at least not in their old form. If no C<binds> |
1356 |
are specified for a node, AnyEvent::MP assumes a default of C<*:*>. |
1357 |
|
1358 |
There are vague plans to implement some form of routing domains, which |
1359 |
might or might not bring back listener-less nodes, but don't count on it. |
1360 |
|
1361 |
The fact that most connections are now optional somewhat mitigates this, |
1362 |
as a node can be effectively unreachable from the outside without any |
1363 |
problems, as long as it isn't a global node and only reaches out to other |
1364 |
nodes (as opposed to being contacted from other nodes). |
1365 |
|
1366 |
=item $AnyEvent::MP::Kernel::WARN has gone. |
1367 |
|
1368 |
AnyEvent has acquired a logging framework (L<AnyEvent::Log>), and AEMP now |
1369 |
uses this, and so should your programs. |
1370 |
|
1371 |
Every module now documents what kinds of messages it generates, with |
1372 |
AnyEvent::MP acting as a catch all. |
1373 |
|
1374 |
On the positive side, this means that instead of setting |
1375 |
C<PERL_ANYEVENT_MP_WARNLEVEL>, you can get away by setting C<AE_VERBOSE> - |
1376 |
much less to type. |
1377 |
|
1378 |
=back |
1379 |
|
1380 |
=head1 LOGGING |
1381 |
|
1382 |
AnyEvent::MP does not normally log anything by itself, but sinc eit is the |
1383 |
root of the contetx hierarchy for AnyEvent::MP modules, it will receive |
1384 |
all log messages by submodules. |
1385 |
|
1386 |
=head1 SEE ALSO |
1387 |
|
1388 |
L<AnyEvent::MP::Intro> - a gentle introduction. |
1389 |
|
1390 |
L<AnyEvent::MP::Kernel> - more, lower-level, stuff. |
1391 |
|
1392 |
L<AnyEvent::MP::Global> - network maintenance and port groups, to find |
1393 |
your applications. |
1394 |
|
1395 |
L<AnyEvent::MP::DataConn> - establish data connections between nodes. |
1396 |
|
1397 |
L<AnyEvent::MP::LogCatcher> - simple service to display log messages from |
1398 |
all nodes. |
1399 |
|
1400 |
L<AnyEvent>. |
1401 |
|
1402 |
=head1 AUTHOR |
1403 |
|
1404 |
Marc Lehmann <schmorp@schmorp.de> |
1405 |
http://home.schmorp.de/ |
1406 |
|
1407 |
=cut |
1408 |
|
1409 |
1 |
1410 |
|