1 |
root |
1.1 |
=head1 NAME |
2 |
|
|
|
3 |
|
|
AnyEvent::MP - multi-processing/message-passing framework |
4 |
|
|
|
5 |
|
|
=head1 SYNOPSIS |
6 |
|
|
|
7 |
|
|
use AnyEvent::MP; |
8 |
|
|
|
9 |
root |
1.22 |
$NODE # contains this node's noderef |
10 |
|
|
NODE # returns this node's noderef |
11 |
|
|
NODE $port # returns the noderef of the port |
12 |
root |
1.2 |
|
13 |
root |
1.38 |
$SELF # receiving/own port id in rcv callbacks |
14 |
|
|
|
15 |
root |
1.48 |
# initialise the node so it can send/receive messages |
16 |
root |
1.63 |
initialise_node; |
17 |
root |
1.48 |
|
18 |
root |
1.38 |
# ports are message endpoints |
19 |
|
|
|
20 |
|
|
# sending messages |
21 |
root |
1.2 |
snd $port, type => data...; |
22 |
root |
1.38 |
snd $port, @msg; |
23 |
|
|
snd @msg_with_first_element_being_a_port; |
24 |
root |
1.2 |
|
25 |
root |
1.50 |
# creating/using ports, the simple way |
26 |
root |
1.53 |
my $simple_port = port { my @msg = @_; 0 }; |
27 |
root |
1.22 |
|
28 |
root |
1.52 |
# creating/using ports, tagged message matching |
29 |
root |
1.38 |
my $port = port; |
30 |
|
|
rcv $port, ping => sub { snd $_[0], "pong"; 0 }; |
31 |
|
|
rcv $port, pong => sub { warn "pong received\n"; 0 }; |
32 |
root |
1.2 |
|
33 |
root |
1.48 |
# create a port on another node |
34 |
|
|
my $port = spawn $node, $initfunc, @initdata; |
35 |
|
|
|
36 |
root |
1.35 |
# monitoring |
37 |
|
|
mon $port, $cb->(@msg) # callback is invoked on death |
38 |
|
|
mon $port, $otherport # kill otherport on abnormal death |
39 |
|
|
mon $port, $otherport, @msg # send message on death |
40 |
|
|
|
41 |
root |
1.45 |
=head1 CURRENT STATUS |
42 |
|
|
|
43 |
root |
1.71 |
bin/aemp - stable. |
44 |
|
|
AnyEvent::MP - stable API, should work. |
45 |
|
|
AnyEvent::MP::Intro - uptodate, but incomplete. |
46 |
|
|
AnyEvent::MP::Kernel - mostly stable. |
47 |
|
|
AnyEvent::MP::Global - stable API, protocol not yet final. |
48 |
root |
1.45 |
|
49 |
|
|
stay tuned. |
50 |
|
|
|
51 |
root |
1.1 |
=head1 DESCRIPTION |
52 |
|
|
|
53 |
root |
1.2 |
This module (-family) implements a simple message passing framework. |
54 |
|
|
|
55 |
|
|
Despite its simplicity, you can securely message other processes running |
56 |
root |
1.67 |
on the same or other hosts, and you can supervise entities remotely. |
57 |
root |
1.2 |
|
58 |
root |
1.23 |
For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
59 |
root |
1.67 |
manual page and the examples under F<eg/>. |
60 |
root |
1.23 |
|
61 |
root |
1.67 |
At the moment, this module family is a bit underdocumented. |
62 |
root |
1.6 |
|
63 |
root |
1.2 |
=head1 CONCEPTS |
64 |
|
|
|
65 |
|
|
=over 4 |
66 |
|
|
|
67 |
|
|
=item port |
68 |
|
|
|
69 |
root |
1.29 |
A port is something you can send messages to (with the C<snd> function). |
70 |
|
|
|
71 |
root |
1.53 |
Ports allow you to register C<rcv> handlers that can match all or just |
72 |
root |
1.64 |
some messages. Messages send to ports will not be queued, regardless of |
73 |
|
|
anything was listening for them or not. |
74 |
root |
1.2 |
|
75 |
root |
1.67 |
=item port ID - C<nodeid#portname> |
76 |
root |
1.2 |
|
77 |
root |
1.67 |
A port ID is the concatenation of a node ID, a hash-mark (C<#>) as |
78 |
|
|
separator, and a port name (a printable string of unspecified format). |
79 |
root |
1.2 |
|
80 |
|
|
=item node |
81 |
|
|
|
82 |
root |
1.53 |
A node is a single process containing at least one port - the node port, |
83 |
root |
1.67 |
which enables nodes to manage each other remotely, and to create new |
84 |
root |
1.53 |
ports. |
85 |
root |
1.2 |
|
86 |
root |
1.67 |
Nodes are either public (have one or more listening ports) or private |
87 |
|
|
(no listening ports). Private nodes cannot talk to other private nodes |
88 |
|
|
currently. |
89 |
root |
1.2 |
|
90 |
root |
1.63 |
=item node ID - C<[a-za-Z0-9_\-.:]+> |
91 |
root |
1.2 |
|
92 |
root |
1.64 |
A node ID is a string that uniquely identifies the node within a |
93 |
|
|
network. Depending on the configuration used, node IDs can look like a |
94 |
|
|
hostname, a hostname and a port, or a random string. AnyEvent::MP itself |
95 |
|
|
doesn't interpret node IDs in any way. |
96 |
|
|
|
97 |
|
|
=item binds - C<ip:port> |
98 |
|
|
|
99 |
|
|
Nodes can only talk to each other by creating some kind of connection to |
100 |
|
|
each other. To do this, nodes should listen on one or more local transport |
101 |
|
|
endpoints - binds. Currently, only standard C<ip:port> specifications can |
102 |
|
|
be used, which specify TCP ports to listen on. |
103 |
|
|
|
104 |
|
|
=item seeds - C<host:port> |
105 |
|
|
|
106 |
|
|
When a node starts, it knows nothing about the network. To teach the node |
107 |
|
|
about the network it first has to contact some other node within the |
108 |
|
|
network. This node is called a seed. |
109 |
|
|
|
110 |
|
|
Seeds are transport endpoint(s) of as many nodes as one wants. Those nodes |
111 |
|
|
are expected to be long-running, and at least one of those should always |
112 |
|
|
be available. When nodes run out of connections (e.g. due to a network |
113 |
|
|
error), they try to re-establish connections to some seednodes again to |
114 |
|
|
join the network. |
115 |
root |
1.29 |
|
116 |
root |
1.67 |
Apart from being sued for seeding, seednodes are not special in any way - |
117 |
|
|
every public node can be a seednode. |
118 |
|
|
|
119 |
root |
1.2 |
=back |
120 |
|
|
|
121 |
root |
1.3 |
=head1 VARIABLES/FUNCTIONS |
122 |
root |
1.2 |
|
123 |
|
|
=over 4 |
124 |
|
|
|
125 |
root |
1.1 |
=cut |
126 |
|
|
|
127 |
|
|
package AnyEvent::MP; |
128 |
|
|
|
129 |
root |
1.44 |
use AnyEvent::MP::Kernel; |
130 |
root |
1.2 |
|
131 |
root |
1.1 |
use common::sense; |
132 |
|
|
|
133 |
root |
1.2 |
use Carp (); |
134 |
|
|
|
135 |
root |
1.1 |
use AE (); |
136 |
|
|
|
137 |
root |
1.2 |
use base "Exporter"; |
138 |
|
|
|
139 |
root |
1.44 |
our $VERSION = $AnyEvent::MP::Kernel::VERSION; |
140 |
root |
1.43 |
|
141 |
root |
1.8 |
our @EXPORT = qw( |
142 |
root |
1.59 |
NODE $NODE *SELF node_of after |
143 |
root |
1.67 |
initialise_node |
144 |
root |
1.61 |
snd rcv mon mon_guard kil reg psub spawn |
145 |
root |
1.22 |
port |
146 |
root |
1.8 |
); |
147 |
root |
1.2 |
|
148 |
root |
1.22 |
our $SELF; |
149 |
|
|
|
150 |
|
|
sub _self_die() { |
151 |
|
|
my $msg = $@; |
152 |
|
|
$msg =~ s/\n+$// unless ref $msg; |
153 |
|
|
kil $SELF, die => $msg; |
154 |
|
|
} |
155 |
|
|
|
156 |
|
|
=item $thisnode = NODE / $NODE |
157 |
|
|
|
158 |
root |
1.67 |
The C<NODE> function returns, and the C<$NODE> variable contains, the node |
159 |
root |
1.64 |
ID of the node running in the current process. This value is initialised by |
160 |
|
|
a call to C<initialise_node>. |
161 |
root |
1.22 |
|
162 |
root |
1.63 |
=item $nodeid = node_of $port |
163 |
root |
1.22 |
|
164 |
root |
1.67 |
Extracts and returns the node ID from a port ID or a node ID. |
165 |
root |
1.34 |
|
166 |
root |
1.69 |
=item initialise_node $profile_name, key => value... |
167 |
root |
1.34 |
|
168 |
root |
1.64 |
Before a node can talk to other nodes on the network (i.e. enter |
169 |
|
|
"distributed mode") it has to initialise itself - the minimum a node needs |
170 |
|
|
to know is its own name, and optionally it should know the addresses of |
171 |
|
|
some other nodes in the network to discover other nodes. |
172 |
root |
1.34 |
|
173 |
|
|
This function initialises a node - it must be called exactly once (or |
174 |
|
|
never) before calling other AnyEvent::MP functions. |
175 |
|
|
|
176 |
root |
1.64 |
The first argument is a profile name. If it is C<undef> or missing, then |
177 |
|
|
the current nodename will be used instead (i.e. F<uname -n>). |
178 |
root |
1.34 |
|
179 |
root |
1.69 |
The function first looks up the profile in the aemp configuration (see the |
180 |
|
|
L<aemp> commandline utility). the profile is calculated as follows: |
181 |
|
|
|
182 |
root |
1.70 |
First, all remaining key => value pairs (all of which are conviniently |
183 |
|
|
undocumented at the moment) will be used. Then they will be overwritten by |
184 |
|
|
any values specified in the global default configuration (see the F<aemp> |
185 |
|
|
utility), then the chain of profiles selected, if any. That means that |
186 |
|
|
the values specified in the profile have highest priority and the values |
187 |
|
|
specified via C<initialise_node> have lowest priority. |
188 |
root |
1.49 |
|
189 |
root |
1.64 |
If the profile specifies a node ID, then this will become the node ID of |
190 |
|
|
this process. If not, then the profile name will be used as node ID. The |
191 |
|
|
special node ID of C<anon/> will be replaced by a random node ID. |
192 |
|
|
|
193 |
|
|
The next step is to look up the binds in the profile, followed by binding |
194 |
|
|
aemp protocol listeners on all binds specified (it is possible and valid |
195 |
|
|
to have no binds, meaning that the node cannot be contacted form the |
196 |
|
|
outside. This means the node cannot talk to other nodes that also have no |
197 |
|
|
binds, but it can still talk to all "normal" nodes). |
198 |
|
|
|
199 |
root |
1.70 |
If the profile does not specify a binds list, then a default of C<*> is |
200 |
|
|
used. |
201 |
root |
1.64 |
|
202 |
|
|
Lastly, the seeds list from the profile is passed to the |
203 |
|
|
L<AnyEvent::MP::Global> module, which will then use it to keep |
204 |
|
|
connectivity with at least on of those seed nodes at any point in time. |
205 |
|
|
|
206 |
|
|
Example: become a distributed node listening on the guessed noderef, or |
207 |
|
|
the one specified via C<aemp> for the current node. This should be the |
208 |
|
|
most common form of invocation for "daemon"-type nodes. |
209 |
root |
1.34 |
|
210 |
|
|
initialise_node; |
211 |
|
|
|
212 |
root |
1.64 |
Example: become an anonymous node. This form is often used for commandline |
213 |
|
|
clients. |
214 |
root |
1.34 |
|
215 |
root |
1.64 |
initialise_node "anon/"; |
216 |
root |
1.34 |
|
217 |
root |
1.64 |
Example: become a distributed node. If there is no profile of the given |
218 |
|
|
name, or no binds list was specified, resolve C<localhost:4044> and bind |
219 |
|
|
on the resulting addresses. |
220 |
root |
1.34 |
|
221 |
root |
1.49 |
initialise_node "localhost:4044"; |
222 |
root |
1.34 |
|
223 |
root |
1.22 |
=item $SELF |
224 |
|
|
|
225 |
|
|
Contains the current port id while executing C<rcv> callbacks or C<psub> |
226 |
|
|
blocks. |
227 |
root |
1.3 |
|
228 |
root |
1.67 |
=item *SELF, SELF, %SELF, @SELF... |
229 |
root |
1.22 |
|
230 |
|
|
Due to some quirks in how perl exports variables, it is impossible to |
231 |
root |
1.67 |
just export C<$SELF>, all the symbols named C<SELF> are exported by this |
232 |
root |
1.22 |
module, but only C<$SELF> is currently used. |
233 |
root |
1.3 |
|
234 |
root |
1.33 |
=item snd $port, type => @data |
235 |
root |
1.3 |
|
236 |
root |
1.33 |
=item snd $port, @msg |
237 |
root |
1.3 |
|
238 |
root |
1.67 |
Send the given message to the given port, which can identify either a |
239 |
|
|
local or a remote port, and must be a port ID. |
240 |
root |
1.8 |
|
241 |
root |
1.67 |
While the message can be almost anything, it is highly recommended to |
242 |
|
|
use a string as first element (a port ID, or some word that indicates a |
243 |
|
|
request type etc.) and to consist if only simple perl values (scalars, |
244 |
|
|
arrays, hashes) - if you think you need to pass an object, think again. |
245 |
|
|
|
246 |
|
|
The message data logically becomes read-only after a call to this |
247 |
|
|
function: modifying any argument (or values referenced by them) is |
248 |
|
|
forbidden, as there can be considerable time between the call to C<snd> |
249 |
|
|
and the time the message is actually being serialised - in fact, it might |
250 |
|
|
never be copied as within the same process it is simply handed to the |
251 |
|
|
receiving port. |
252 |
root |
1.3 |
|
253 |
|
|
The type of data you can transfer depends on the transport protocol: when |
254 |
|
|
JSON is used, then only strings, numbers and arrays and hashes consisting |
255 |
|
|
of those are allowed (no objects). When Storable is used, then anything |
256 |
|
|
that Storable can serialise and deserialise is allowed, and for the local |
257 |
root |
1.67 |
node, anything can be passed. Best rely only on the common denominator of |
258 |
|
|
these. |
259 |
root |
1.3 |
|
260 |
root |
1.22 |
=item $local_port = port |
261 |
root |
1.2 |
|
262 |
root |
1.50 |
Create a new local port object and returns its port ID. Initially it has |
263 |
|
|
no callbacks set and will throw an error when it receives messages. |
264 |
root |
1.10 |
|
265 |
root |
1.50 |
=item $local_port = port { my @msg = @_ } |
266 |
root |
1.15 |
|
267 |
root |
1.50 |
Creates a new local port, and returns its ID. Semantically the same as |
268 |
|
|
creating a port and calling C<rcv $port, $callback> on it. |
269 |
root |
1.15 |
|
270 |
root |
1.50 |
The block will be called for every message received on the port, with the |
271 |
|
|
global variable C<$SELF> set to the port ID. Runtime errors will cause the |
272 |
|
|
port to be C<kil>ed. The message will be passed as-is, no extra argument |
273 |
|
|
(i.e. no port ID) will be passed to the callback. |
274 |
root |
1.15 |
|
275 |
root |
1.50 |
If you want to stop/destroy the port, simply C<kil> it: |
276 |
root |
1.15 |
|
277 |
root |
1.50 |
my $port = port { |
278 |
|
|
my @msg = @_; |
279 |
|
|
... |
280 |
|
|
kil $SELF; |
281 |
root |
1.15 |
}; |
282 |
root |
1.10 |
|
283 |
|
|
=cut |
284 |
|
|
|
285 |
root |
1.33 |
sub rcv($@); |
286 |
|
|
|
287 |
root |
1.50 |
sub _kilme { |
288 |
|
|
die "received message on port without callback"; |
289 |
|
|
} |
290 |
|
|
|
291 |
root |
1.22 |
sub port(;&) { |
292 |
|
|
my $id = "$UNIQ." . $ID++; |
293 |
|
|
my $port = "$NODE#$id"; |
294 |
|
|
|
295 |
root |
1.50 |
rcv $port, shift || \&_kilme; |
296 |
root |
1.10 |
|
297 |
root |
1.22 |
$port |
298 |
root |
1.10 |
} |
299 |
|
|
|
300 |
root |
1.50 |
=item rcv $local_port, $callback->(@msg) |
301 |
root |
1.31 |
|
302 |
root |
1.50 |
Replaces the default callback on the specified port. There is no way to |
303 |
|
|
remove the default callback: use C<sub { }> to disable it, or better |
304 |
|
|
C<kil> the port when it is no longer needed. |
305 |
root |
1.3 |
|
306 |
root |
1.33 |
The global C<$SELF> (exported by this module) contains C<$port> while |
307 |
root |
1.50 |
executing the callback. Runtime errors during callback execution will |
308 |
|
|
result in the port being C<kil>ed. |
309 |
root |
1.22 |
|
310 |
root |
1.50 |
The default callback received all messages not matched by a more specific |
311 |
|
|
C<tag> match. |
312 |
root |
1.22 |
|
313 |
root |
1.50 |
=item rcv $local_port, tag => $callback->(@msg_without_tag), ... |
314 |
root |
1.3 |
|
315 |
root |
1.54 |
Register (or replace) callbacks to be called on messages starting with the |
316 |
|
|
given tag on the given port (and return the port), or unregister it (when |
317 |
|
|
C<$callback> is C<$undef> or missing). There can only be one callback |
318 |
|
|
registered for each tag. |
319 |
root |
1.3 |
|
320 |
root |
1.50 |
The original message will be passed to the callback, after the first |
321 |
|
|
element (the tag) has been removed. The callback will use the same |
322 |
|
|
environment as the default callback (see above). |
323 |
root |
1.3 |
|
324 |
root |
1.36 |
Example: create a port and bind receivers on it in one go. |
325 |
|
|
|
326 |
|
|
my $port = rcv port, |
327 |
root |
1.50 |
msg1 => sub { ... }, |
328 |
|
|
msg2 => sub { ... }, |
329 |
root |
1.36 |
; |
330 |
|
|
|
331 |
|
|
Example: create a port, bind receivers and send it in a message elsewhere |
332 |
|
|
in one go: |
333 |
|
|
|
334 |
|
|
snd $otherport, reply => |
335 |
|
|
rcv port, |
336 |
root |
1.50 |
msg1 => sub { ... }, |
337 |
root |
1.36 |
... |
338 |
|
|
; |
339 |
|
|
|
340 |
root |
1.54 |
Example: temporarily register a rcv callback for a tag matching some port |
341 |
|
|
(e.g. for a rpc reply) and unregister it after a message was received. |
342 |
|
|
|
343 |
|
|
rcv $port, $otherport => sub { |
344 |
|
|
my @reply = @_; |
345 |
|
|
|
346 |
|
|
rcv $SELF, $otherport; |
347 |
|
|
}; |
348 |
|
|
|
349 |
root |
1.3 |
=cut |
350 |
|
|
|
351 |
|
|
sub rcv($@) { |
352 |
root |
1.33 |
my $port = shift; |
353 |
|
|
my ($noderef, $portid) = split /#/, $port, 2; |
354 |
root |
1.3 |
|
355 |
root |
1.58 |
$NODE{$noderef} == $NODE{""} |
356 |
root |
1.33 |
or Carp::croak "$port: rcv can only be called on local ports, caught"; |
357 |
root |
1.22 |
|
358 |
root |
1.50 |
while (@_) { |
359 |
|
|
if (ref $_[0]) { |
360 |
|
|
if (my $self = $PORT_DATA{$portid}) { |
361 |
|
|
"AnyEvent::MP::Port" eq ref $self |
362 |
|
|
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
363 |
root |
1.33 |
|
364 |
root |
1.50 |
$self->[2] = shift; |
365 |
|
|
} else { |
366 |
|
|
my $cb = shift; |
367 |
|
|
$PORT{$portid} = sub { |
368 |
|
|
local $SELF = $port; |
369 |
|
|
eval { &$cb }; _self_die if $@; |
370 |
|
|
}; |
371 |
|
|
} |
372 |
|
|
} elsif (defined $_[0]) { |
373 |
|
|
my $self = $PORT_DATA{$portid} ||= do { |
374 |
|
|
my $self = bless [$PORT{$port} || sub { }, { }, $port], "AnyEvent::MP::Port"; |
375 |
|
|
|
376 |
|
|
$PORT{$portid} = sub { |
377 |
|
|
local $SELF = $port; |
378 |
|
|
|
379 |
|
|
if (my $cb = $self->[1]{$_[0]}) { |
380 |
|
|
shift; |
381 |
|
|
eval { &$cb }; _self_die if $@; |
382 |
|
|
} else { |
383 |
|
|
&{ $self->[0] }; |
384 |
root |
1.33 |
} |
385 |
|
|
}; |
386 |
root |
1.50 |
|
387 |
|
|
$self |
388 |
root |
1.33 |
}; |
389 |
|
|
|
390 |
root |
1.50 |
"AnyEvent::MP::Port" eq ref $self |
391 |
|
|
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
392 |
root |
1.22 |
|
393 |
root |
1.50 |
my ($tag, $cb) = splice @_, 0, 2; |
394 |
root |
1.33 |
|
395 |
root |
1.50 |
if (defined $cb) { |
396 |
|
|
$self->[1]{$tag} = $cb; |
397 |
root |
1.33 |
} else { |
398 |
root |
1.50 |
delete $self->[1]{$tag}; |
399 |
root |
1.33 |
} |
400 |
root |
1.22 |
} |
401 |
root |
1.3 |
} |
402 |
root |
1.31 |
|
403 |
root |
1.33 |
$port |
404 |
root |
1.2 |
} |
405 |
|
|
|
406 |
root |
1.22 |
=item $closure = psub { BLOCK } |
407 |
root |
1.2 |
|
408 |
root |
1.22 |
Remembers C<$SELF> and creates a closure out of the BLOCK. When the |
409 |
|
|
closure is executed, sets up the environment in the same way as in C<rcv> |
410 |
|
|
callbacks, i.e. runtime errors will cause the port to get C<kil>ed. |
411 |
|
|
|
412 |
|
|
This is useful when you register callbacks from C<rcv> callbacks: |
413 |
|
|
|
414 |
|
|
rcv delayed_reply => sub { |
415 |
|
|
my ($delay, @reply) = @_; |
416 |
|
|
my $timer = AE::timer $delay, 0, psub { |
417 |
|
|
snd @reply, $SELF; |
418 |
|
|
}; |
419 |
|
|
}; |
420 |
root |
1.3 |
|
421 |
root |
1.8 |
=cut |
422 |
root |
1.3 |
|
423 |
root |
1.22 |
sub psub(&) { |
424 |
|
|
my $cb = shift; |
425 |
root |
1.3 |
|
426 |
root |
1.22 |
my $port = $SELF |
427 |
|
|
or Carp::croak "psub can only be called from within rcv or psub callbacks, not"; |
428 |
root |
1.1 |
|
429 |
root |
1.22 |
sub { |
430 |
|
|
local $SELF = $port; |
431 |
root |
1.2 |
|
432 |
root |
1.22 |
if (wantarray) { |
433 |
|
|
my @res = eval { &$cb }; |
434 |
|
|
_self_die if $@; |
435 |
|
|
@res |
436 |
|
|
} else { |
437 |
|
|
my $res = eval { &$cb }; |
438 |
|
|
_self_die if $@; |
439 |
|
|
$res |
440 |
|
|
} |
441 |
|
|
} |
442 |
root |
1.2 |
} |
443 |
|
|
|
444 |
root |
1.67 |
=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies |
445 |
root |
1.32 |
|
446 |
root |
1.67 |
=item $guard = mon $port, $rcvport # kill $rcvport when $port dies |
447 |
root |
1.36 |
|
448 |
root |
1.67 |
=item $guard = mon $port # kill $SELF when $port dies |
449 |
root |
1.32 |
|
450 |
root |
1.67 |
=item $guard = mon $port, $rcvport, @msg # send a message when $port dies |
451 |
root |
1.32 |
|
452 |
root |
1.42 |
Monitor the given port and do something when the port is killed or |
453 |
|
|
messages to it were lost, and optionally return a guard that can be used |
454 |
|
|
to stop monitoring again. |
455 |
|
|
|
456 |
|
|
C<mon> effectively guarantees that, in the absence of hardware failures, |
457 |
root |
1.67 |
after starting the monitor, either all messages sent to the port will |
458 |
|
|
arrive, or the monitoring action will be invoked after possible message |
459 |
|
|
loss has been detected. No messages will be lost "in between" (after |
460 |
|
|
the first lost message no further messages will be received by the |
461 |
root |
1.42 |
port). After the monitoring action was invoked, further messages might get |
462 |
|
|
delivered again. |
463 |
root |
1.32 |
|
464 |
root |
1.67 |
Note that monitoring-actions are one-shot: once messages are lost (and a |
465 |
|
|
monitoring alert was raised), they are removed and will not trigger again. |
466 |
root |
1.58 |
|
467 |
root |
1.36 |
In the first form (callback), the callback is simply called with any |
468 |
|
|
number of C<@reason> elements (no @reason means that the port was deleted |
469 |
root |
1.32 |
"normally"). Note also that I<< the callback B<must> never die >>, so use |
470 |
|
|
C<eval> if unsure. |
471 |
|
|
|
472 |
root |
1.43 |
In the second form (another port given), the other port (C<$rcvport>) |
473 |
root |
1.36 |
will be C<kil>'ed with C<@reason>, iff a @reason was specified, i.e. on |
474 |
|
|
"normal" kils nothing happens, while under all other conditions, the other |
475 |
|
|
port is killed with the same reason. |
476 |
root |
1.32 |
|
477 |
root |
1.36 |
The third form (kill self) is the same as the second form, except that |
478 |
|
|
C<$rvport> defaults to C<$SELF>. |
479 |
|
|
|
480 |
|
|
In the last form (message), a message of the form C<@msg, @reason> will be |
481 |
|
|
C<snd>. |
482 |
root |
1.32 |
|
483 |
root |
1.37 |
As a rule of thumb, monitoring requests should always monitor a port from |
484 |
|
|
a local port (or callback). The reason is that kill messages might get |
485 |
|
|
lost, just like any other message. Another less obvious reason is that |
486 |
|
|
even monitoring requests can get lost (for exmaple, when the connection |
487 |
|
|
to the other node goes down permanently). When monitoring a port locally |
488 |
|
|
these problems do not exist. |
489 |
|
|
|
490 |
root |
1.32 |
Example: call a given callback when C<$port> is killed. |
491 |
|
|
|
492 |
|
|
mon $port, sub { warn "port died because of <@_>\n" }; |
493 |
|
|
|
494 |
|
|
Example: kill ourselves when C<$port> is killed abnormally. |
495 |
|
|
|
496 |
root |
1.36 |
mon $port; |
497 |
root |
1.32 |
|
498 |
root |
1.36 |
Example: send us a restart message when another C<$port> is killed. |
499 |
root |
1.32 |
|
500 |
|
|
mon $port, $self => "restart"; |
501 |
|
|
|
502 |
|
|
=cut |
503 |
|
|
|
504 |
|
|
sub mon { |
505 |
|
|
my ($noderef, $port) = split /#/, shift, 2; |
506 |
|
|
|
507 |
|
|
my $node = $NODE{$noderef} || add_node $noderef; |
508 |
|
|
|
509 |
root |
1.41 |
my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,'; |
510 |
root |
1.32 |
|
511 |
|
|
unless (ref $cb) { |
512 |
|
|
if (@_) { |
513 |
|
|
# send a kill info message |
514 |
root |
1.41 |
my (@msg) = ($cb, @_); |
515 |
root |
1.32 |
$cb = sub { snd @msg, @_ }; |
516 |
|
|
} else { |
517 |
|
|
# simply kill other port |
518 |
|
|
my $port = $cb; |
519 |
|
|
$cb = sub { kil $port, @_ if @_ }; |
520 |
|
|
} |
521 |
|
|
} |
522 |
|
|
|
523 |
|
|
$node->monitor ($port, $cb); |
524 |
|
|
|
525 |
|
|
defined wantarray |
526 |
|
|
and AnyEvent::Util::guard { $node->unmonitor ($port, $cb) } |
527 |
|
|
} |
528 |
|
|
|
529 |
|
|
=item $guard = mon_guard $port, $ref, $ref... |
530 |
|
|
|
531 |
|
|
Monitors the given C<$port> and keeps the passed references. When the port |
532 |
|
|
is killed, the references will be freed. |
533 |
|
|
|
534 |
|
|
Optionally returns a guard that will stop the monitoring. |
535 |
|
|
|
536 |
|
|
This function is useful when you create e.g. timers or other watchers and |
537 |
root |
1.67 |
want to free them when the port gets killed (note the use of C<psub>): |
538 |
root |
1.32 |
|
539 |
|
|
$port->rcv (start => sub { |
540 |
root |
1.67 |
my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub { |
541 |
root |
1.32 |
undef $timer if 0.9 < rand; |
542 |
|
|
}); |
543 |
|
|
}); |
544 |
|
|
|
545 |
|
|
=cut |
546 |
|
|
|
547 |
|
|
sub mon_guard { |
548 |
|
|
my ($port, @refs) = @_; |
549 |
|
|
|
550 |
root |
1.36 |
#TODO: mon-less form? |
551 |
|
|
|
552 |
root |
1.32 |
mon $port, sub { 0 && @refs } |
553 |
|
|
} |
554 |
|
|
|
555 |
root |
1.33 |
=item kil $port[, @reason] |
556 |
root |
1.32 |
|
557 |
|
|
Kill the specified port with the given C<@reason>. |
558 |
|
|
|
559 |
root |
1.67 |
If no C<@reason> is specified, then the port is killed "normally" (ports |
560 |
|
|
monitoring other ports will not necessarily die because a port dies |
561 |
|
|
"normally"). |
562 |
root |
1.32 |
|
563 |
|
|
Otherwise, linked ports get killed with the same reason (second form of |
564 |
root |
1.67 |
C<mon>, see above). |
565 |
root |
1.32 |
|
566 |
|
|
Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks |
567 |
|
|
will be reported as reason C<< die => $@ >>. |
568 |
|
|
|
569 |
|
|
Transport/communication errors are reported as C<< transport_error => |
570 |
|
|
$message >>. |
571 |
|
|
|
572 |
root |
1.38 |
=cut |
573 |
|
|
|
574 |
|
|
=item $port = spawn $node, $initfunc[, @initdata] |
575 |
|
|
|
576 |
|
|
Creates a port on the node C<$node> (which can also be a port ID, in which |
577 |
|
|
case it's the node where that port resides). |
578 |
|
|
|
579 |
root |
1.67 |
The port ID of the newly created port is returned immediately, and it is |
580 |
|
|
possible to immediately start sending messages or to monitor the port. |
581 |
root |
1.38 |
|
582 |
root |
1.67 |
After the port has been created, the init function is called on the remote |
583 |
|
|
node, in the same context as a C<rcv> callback. This function must be a |
584 |
|
|
fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To |
585 |
|
|
specify a function in the main program, use C<::name>. |
586 |
root |
1.38 |
|
587 |
|
|
If the function doesn't exist, then the node tries to C<require> |
588 |
|
|
the package, then the package above the package and so on (e.g. |
589 |
|
|
C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
590 |
|
|
exists or it runs out of package names. |
591 |
|
|
|
592 |
|
|
The init function is then called with the newly-created port as context |
593 |
|
|
object (C<$SELF>) and the C<@initdata> values as arguments. |
594 |
|
|
|
595 |
root |
1.67 |
A common idiom is to pass a local port, immediately monitor the spawned |
596 |
|
|
port, and in the remote init function, immediately monitor the passed |
597 |
|
|
local port. This two-way monitoring ensures that both ports get cleaned up |
598 |
|
|
when there is a problem. |
599 |
root |
1.38 |
|
600 |
|
|
Example: spawn a chat server port on C<$othernode>. |
601 |
|
|
|
602 |
|
|
# this node, executed from within a port context: |
603 |
|
|
my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
604 |
|
|
mon $server; |
605 |
|
|
|
606 |
|
|
# init function on C<$othernode> |
607 |
|
|
sub connect { |
608 |
|
|
my ($srcport) = @_; |
609 |
|
|
|
610 |
|
|
mon $srcport; |
611 |
|
|
|
612 |
|
|
rcv $SELF, sub { |
613 |
|
|
... |
614 |
|
|
}; |
615 |
|
|
} |
616 |
|
|
|
617 |
|
|
=cut |
618 |
|
|
|
619 |
|
|
sub _spawn { |
620 |
|
|
my $port = shift; |
621 |
|
|
my $init = shift; |
622 |
|
|
|
623 |
|
|
local $SELF = "$NODE#$port"; |
624 |
|
|
eval { |
625 |
|
|
&{ load_func $init } |
626 |
|
|
}; |
627 |
|
|
_self_die if $@; |
628 |
|
|
} |
629 |
|
|
|
630 |
|
|
sub spawn(@) { |
631 |
|
|
my ($noderef, undef) = split /#/, shift, 2; |
632 |
|
|
|
633 |
|
|
my $id = "$RUNIQ." . $ID++; |
634 |
|
|
|
635 |
root |
1.39 |
$_[0] =~ /::/ |
636 |
|
|
or Carp::croak "spawn init function must be a fully-qualified name, caught"; |
637 |
|
|
|
638 |
root |
1.55 |
snd_to_func $noderef, "AnyEvent::MP::_spawn" => $id, @_; |
639 |
root |
1.38 |
|
640 |
|
|
"$noderef#$id" |
641 |
|
|
} |
642 |
|
|
|
643 |
root |
1.59 |
=item after $timeout, @msg |
644 |
|
|
|
645 |
|
|
=item after $timeout, $callback |
646 |
|
|
|
647 |
|
|
Either sends the given message, or call the given callback, after the |
648 |
|
|
specified number of seconds. |
649 |
|
|
|
650 |
root |
1.67 |
This is simply a utility function that comes in handy at times - the |
651 |
|
|
AnyEvent::MP author is not convinced of the wisdom of having it, though, |
652 |
|
|
so it may go away in the future. |
653 |
root |
1.59 |
|
654 |
|
|
=cut |
655 |
|
|
|
656 |
|
|
sub after($@) { |
657 |
|
|
my ($timeout, @action) = @_; |
658 |
|
|
|
659 |
|
|
my $t; $t = AE::timer $timeout, 0, sub { |
660 |
|
|
undef $t; |
661 |
|
|
ref $action[0] |
662 |
|
|
? $action[0]() |
663 |
|
|
: snd @action; |
664 |
|
|
}; |
665 |
|
|
} |
666 |
|
|
|
667 |
root |
1.8 |
=back |
668 |
|
|
|
669 |
root |
1.26 |
=head1 AnyEvent::MP vs. Distributed Erlang |
670 |
|
|
|
671 |
root |
1.35 |
AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
672 |
|
|
== aemp node, Erlang process == aemp port), so many of the documents and |
673 |
|
|
programming techniques employed by Erlang apply to AnyEvent::MP. Here is a |
674 |
root |
1.27 |
sample: |
675 |
|
|
|
676 |
root |
1.35 |
http://www.Erlang.se/doc/programming_rules.shtml |
677 |
|
|
http://Erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4 |
678 |
|
|
http://Erlang.org/download/Erlang-book-part1.pdf # chapters 5 and 6 |
679 |
|
|
http://Erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5 |
680 |
root |
1.27 |
|
681 |
|
|
Despite the similarities, there are also some important differences: |
682 |
root |
1.26 |
|
683 |
|
|
=over 4 |
684 |
|
|
|
685 |
root |
1.65 |
=item * Node IDs are arbitrary strings in AEMP. |
686 |
root |
1.26 |
|
687 |
root |
1.65 |
Erlang relies on special naming and DNS to work everywhere in the same |
688 |
|
|
way. AEMP relies on each node somehow knowing its own address(es) (e.g. by |
689 |
|
|
configuraiton or DNS), but will otherwise discover other odes itself. |
690 |
root |
1.27 |
|
691 |
root |
1.54 |
=item * Erlang has a "remote ports are like local ports" philosophy, AEMP |
692 |
root |
1.51 |
uses "local ports are like remote ports". |
693 |
|
|
|
694 |
|
|
The failure modes for local ports are quite different (runtime errors |
695 |
|
|
only) then for remote ports - when a local port dies, you I<know> it dies, |
696 |
|
|
when a connection to another node dies, you know nothing about the other |
697 |
|
|
port. |
698 |
|
|
|
699 |
|
|
Erlang pretends remote ports are as reliable as local ports, even when |
700 |
|
|
they are not. |
701 |
|
|
|
702 |
|
|
AEMP encourages a "treat remote ports differently" philosophy, with local |
703 |
|
|
ports being the special case/exception, where transport errors cannot |
704 |
|
|
occur. |
705 |
|
|
|
706 |
root |
1.26 |
=item * Erlang uses processes and a mailbox, AEMP does not queue. |
707 |
|
|
|
708 |
root |
1.51 |
Erlang uses processes that selectively receive messages, and therefore |
709 |
|
|
needs a queue. AEMP is event based, queuing messages would serve no |
710 |
|
|
useful purpose. For the same reason the pattern-matching abilities of |
711 |
|
|
AnyEvent::MP are more limited, as there is little need to be able to |
712 |
|
|
filter messages without dequeing them. |
713 |
root |
1.26 |
|
714 |
root |
1.35 |
(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP). |
715 |
root |
1.26 |
|
716 |
|
|
=item * Erlang sends are synchronous, AEMP sends are asynchronous. |
717 |
|
|
|
718 |
root |
1.51 |
Sending messages in Erlang is synchronous and blocks the process (and |
719 |
|
|
so does not need a queue that can overflow). AEMP sends are immediate, |
720 |
|
|
connection establishment is handled in the background. |
721 |
root |
1.26 |
|
722 |
root |
1.51 |
=item * Erlang suffers from silent message loss, AEMP does not. |
723 |
root |
1.26 |
|
724 |
|
|
Erlang makes few guarantees on messages delivery - messages can get lost |
725 |
|
|
without any of the processes realising it (i.e. you send messages a, b, |
726 |
|
|
and c, and the other side only receives messages a and c). |
727 |
|
|
|
728 |
root |
1.66 |
AEMP guarantees correct ordering, and the guarantee that after one message |
729 |
|
|
is lost, all following ones sent to the same port are lost as well, until |
730 |
|
|
monitoring raises an error, so there are no silent "holes" in the message |
731 |
|
|
sequence. |
732 |
root |
1.26 |
|
733 |
|
|
=item * Erlang can send messages to the wrong port, AEMP does not. |
734 |
|
|
|
735 |
root |
1.51 |
In Erlang it is quite likely that a node that restarts reuses a process ID |
736 |
|
|
known to other nodes for a completely different process, causing messages |
737 |
|
|
destined for that process to end up in an unrelated process. |
738 |
root |
1.26 |
|
739 |
|
|
AEMP never reuses port IDs, so old messages or old port IDs floating |
740 |
|
|
around in the network will not be sent to an unrelated port. |
741 |
|
|
|
742 |
|
|
=item * Erlang uses unprotected connections, AEMP uses secure |
743 |
|
|
authentication and can use TLS. |
744 |
|
|
|
745 |
root |
1.66 |
AEMP can use a proven protocol - TLS - to protect connections and |
746 |
root |
1.26 |
securely authenticate nodes. |
747 |
|
|
|
748 |
root |
1.28 |
=item * The AEMP protocol is optimised for both text-based and binary |
749 |
|
|
communications. |
750 |
|
|
|
751 |
root |
1.66 |
The AEMP protocol, unlike the Erlang protocol, supports both programming |
752 |
|
|
language independent text-only protocols (good for debugging) and binary, |
753 |
root |
1.67 |
language-specific serialisers (e.g. Storable). By default, unless TLS is |
754 |
|
|
used, the protocol is actually completely text-based. |
755 |
root |
1.28 |
|
756 |
|
|
It has also been carefully designed to be implementable in other languages |
757 |
root |
1.66 |
with a minimum of work while gracefully degrading functionality to make the |
758 |
root |
1.28 |
protocol simple. |
759 |
|
|
|
760 |
root |
1.35 |
=item * AEMP has more flexible monitoring options than Erlang. |
761 |
|
|
|
762 |
|
|
In Erlang, you can chose to receive I<all> exit signals as messages |
763 |
|
|
or I<none>, there is no in-between, so monitoring single processes is |
764 |
|
|
difficult to implement. Monitoring in AEMP is more flexible than in |
765 |
|
|
Erlang, as one can choose between automatic kill, exit message or callback |
766 |
|
|
on a per-process basis. |
767 |
|
|
|
768 |
root |
1.37 |
=item * Erlang tries to hide remote/local connections, AEMP does not. |
769 |
root |
1.35 |
|
770 |
root |
1.67 |
Monitoring in Erlang is not an indicator of process death/crashes, in the |
771 |
|
|
same way as linking is (except linking is unreliable in Erlang). |
772 |
root |
1.37 |
|
773 |
|
|
In AEMP, you don't "look up" registered port names or send to named ports |
774 |
|
|
that might or might not be persistent. Instead, you normally spawn a port |
775 |
root |
1.67 |
on the remote node. The init function monitors you, and you monitor the |
776 |
|
|
remote port. Since both monitors are local to the node, they are much more |
777 |
|
|
reliable (no need for C<spawn_link>). |
778 |
root |
1.37 |
|
779 |
|
|
This also saves round-trips and avoids sending messages to the wrong port |
780 |
|
|
(hard to do in Erlang). |
781 |
root |
1.35 |
|
782 |
root |
1.26 |
=back |
783 |
|
|
|
784 |
root |
1.46 |
=head1 RATIONALE |
785 |
|
|
|
786 |
|
|
=over 4 |
787 |
|
|
|
788 |
root |
1.67 |
=item Why strings for port and node IDs, why not objects? |
789 |
root |
1.46 |
|
790 |
|
|
We considered "objects", but found that the actual number of methods |
791 |
root |
1.67 |
that can be called are quite low. Since port and node IDs travel over |
792 |
root |
1.46 |
the network frequently, the serialising/deserialising would add lots of |
793 |
root |
1.67 |
overhead, as well as having to keep a proxy object everywhere. |
794 |
root |
1.46 |
|
795 |
|
|
Strings can easily be printed, easily serialised etc. and need no special |
796 |
|
|
procedures to be "valid". |
797 |
|
|
|
798 |
root |
1.67 |
And as a result, a miniport consists of a single closure stored in a |
799 |
|
|
global hash - it can't become much cheaper. |
800 |
root |
1.47 |
|
801 |
root |
1.67 |
=item Why favour JSON, why not a real serialising format such as Storable? |
802 |
root |
1.46 |
|
803 |
|
|
In fact, any AnyEvent::MP node will happily accept Storable as framing |
804 |
|
|
format, but currently there is no way to make a node use Storable by |
805 |
root |
1.67 |
default (although all nodes will accept it). |
806 |
root |
1.46 |
|
807 |
|
|
The default framing protocol is JSON because a) JSON::XS is many times |
808 |
|
|
faster for small messages and b) most importantly, after years of |
809 |
|
|
experience we found that object serialisation is causing more problems |
810 |
root |
1.67 |
than it solves: Just like function calls, objects simply do not travel |
811 |
root |
1.46 |
easily over the network, mostly because they will always be a copy, so you |
812 |
|
|
always have to re-think your design. |
813 |
|
|
|
814 |
|
|
Keeping your messages simple, concentrating on data structures rather than |
815 |
|
|
objects, will keep your messages clean, tidy and efficient. |
816 |
|
|
|
817 |
|
|
=back |
818 |
|
|
|
819 |
root |
1.1 |
=head1 SEE ALSO |
820 |
|
|
|
821 |
root |
1.68 |
L<AnyEvent::MP::Intro> - a gentle introduction. |
822 |
|
|
|
823 |
|
|
L<AnyEvent::MP::Kernel> - more, lower-level, stuff. |
824 |
|
|
|
825 |
|
|
L<AnyEvent::MP::Global> - network maintainance and port groups, to find |
826 |
|
|
your applications. |
827 |
|
|
|
828 |
root |
1.1 |
L<AnyEvent>. |
829 |
|
|
|
830 |
|
|
=head1 AUTHOR |
831 |
|
|
|
832 |
|
|
Marc Lehmann <schmorp@schmorp.de> |
833 |
|
|
http://home.schmorp.de/ |
834 |
|
|
|
835 |
|
|
=cut |
836 |
|
|
|
837 |
|
|
1 |
838 |
|
|
|