… | |
… | |
39 | |
39 | |
40 | =head1 CURRENT STATUS |
40 | =head1 CURRENT STATUS |
41 | |
41 | |
42 | bin/aemp - stable. |
42 | bin/aemp - stable. |
43 | AnyEvent::MP - stable API, should work. |
43 | AnyEvent::MP - stable API, should work. |
44 | AnyEvent::MP::Intro - uptodate, but incomplete. |
44 | AnyEvent::MP::Intro - explains most concepts. |
45 | AnyEvent::MP::Kernel - mostly stable. |
45 | AnyEvent::MP::Kernel - mostly stable. |
46 | AnyEvent::MP::Global - stable API, protocol not yet final. |
46 | AnyEvent::MP::Global - stable but incomplete, protocol not yet final. |
47 | |
47 | |
48 | stay tuned. |
48 | stay tuned. |
49 | |
49 | |
50 | =head1 DESCRIPTION |
50 | =head1 DESCRIPTION |
51 | |
51 | |
52 | This module (-family) implements a simple message passing framework. |
52 | This module (-family) implements a simple message passing framework. |
53 | |
53 | |
… | |
… | |
55 | on the same or other hosts, and you can supervise entities remotely. |
55 | on the same or other hosts, and you can supervise entities remotely. |
56 | |
56 | |
57 | For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
57 | For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
58 | manual page and the examples under F<eg/>. |
58 | manual page and the examples under F<eg/>. |
59 | |
59 | |
60 | At the moment, this module family is a bit underdocumented. |
|
|
61 | |
|
|
62 | =head1 CONCEPTS |
60 | =head1 CONCEPTS |
63 | |
61 | |
64 | =over 4 |
62 | =over 4 |
65 | |
63 | |
66 | =item port |
64 | =item port |
67 | |
65 | |
68 | A port is something you can send messages to (with the C<snd> function). |
66 | Not to be confused with a TCP port, a "port" is something you can send |
|
|
67 | messages to (with the C<snd> function). |
69 | |
68 | |
70 | Ports allow you to register C<rcv> handlers that can match all or just |
69 | Ports allow you to register C<rcv> handlers that can match all or just |
71 | some messages. Messages send to ports will not be queued, regardless of |
70 | some messages. Messages send to ports will not be queued, regardless of |
72 | anything was listening for them or not. |
71 | anything was listening for them or not. |
73 | |
72 | |
… | |
… | |
84 | |
83 | |
85 | Nodes are either public (have one or more listening ports) or private |
84 | Nodes are either public (have one or more listening ports) or private |
86 | (no listening ports). Private nodes cannot talk to other private nodes |
85 | (no listening ports). Private nodes cannot talk to other private nodes |
87 | currently. |
86 | currently. |
88 | |
87 | |
89 | =item node ID - C<[a-za-Z0-9_\-.:]+> |
88 | =item node ID - C<[A-Z_][a-zA-Z0-9_\-.:]*> |
90 | |
89 | |
91 | A node ID is a string that uniquely identifies the node within a |
90 | A node ID is a string that uniquely identifies the node within a |
92 | network. Depending on the configuration used, node IDs can look like a |
91 | network. Depending on the configuration used, node IDs can look like a |
93 | hostname, a hostname and a port, or a random string. AnyEvent::MP itself |
92 | hostname, a hostname and a port, or a random string. AnyEvent::MP itself |
94 | doesn't interpret node IDs in any way. |
93 | doesn't interpret node IDs in any way. |
… | |
… | |
98 | Nodes can only talk to each other by creating some kind of connection to |
97 | Nodes can only talk to each other by creating some kind of connection to |
99 | each other. To do this, nodes should listen on one or more local transport |
98 | each other. To do this, nodes should listen on one or more local transport |
100 | endpoints - binds. Currently, only standard C<ip:port> specifications can |
99 | endpoints - binds. Currently, only standard C<ip:port> specifications can |
101 | be used, which specify TCP ports to listen on. |
100 | be used, which specify TCP ports to listen on. |
102 | |
101 | |
103 | =item seeds - C<host:port> |
102 | =item seed nodes |
104 | |
103 | |
105 | When a node starts, it knows nothing about the network. To teach the node |
104 | When a node starts, it knows nothing about the network. To teach the node |
106 | about the network it first has to contact some other node within the |
105 | about the network it first has to contact some other node within the |
107 | network. This node is called a seed. |
106 | network. This node is called a seed. |
108 | |
107 | |
109 | Seeds are transport endpoint(s) of as many nodes as one wants. Those nodes |
108 | Apart from the fact that other nodes know them as seed nodes and they have |
|
|
109 | to have fixed listening addresses, seed nodes are perfectly normal nodes - |
|
|
110 | any node can function as a seed node for others. |
|
|
111 | |
|
|
112 | In addition to discovering the network, seed nodes are also used to |
|
|
113 | maintain the network and to connect nodes that otherwise would have |
|
|
114 | trouble connecting. They form the backbone of an AnyEvent::MP network. |
|
|
115 | |
110 | are expected to be long-running, and at least one of those should always |
116 | Seed nodes are expected to be long-running, and at least one seed node |
111 | be available. When nodes run out of connections (e.g. due to a network |
117 | should always be available. They should also be relatively responsive - a |
112 | error), they try to re-establish connections to some seednodes again to |
118 | seed node that blocks for long periods will slow down everybody else. |
113 | join the network. |
|
|
114 | |
119 | |
115 | Apart from being sued for seeding, seednodes are not special in any way - |
120 | =item seeds - C<host:port> |
116 | every public node can be a seednode. |
121 | |
|
|
122 | Seeds are transport endpoint(s) (usually a hostname/IP address and a |
|
|
123 | TCP port) of nodes thta should be used as seed nodes. |
|
|
124 | |
|
|
125 | The nodes listening on those endpoints are expected to be long-running, |
|
|
126 | and at least one of those should always be available. When nodes run out |
|
|
127 | of connections (e.g. due to a network error), they try to re-establish |
|
|
128 | connections to some seednodes again to join the network. |
117 | |
129 | |
118 | =back |
130 | =back |
119 | |
131 | |
120 | =head1 VARIABLES/FUNCTIONS |
132 | =head1 VARIABLES/FUNCTIONS |
121 | |
133 | |
… | |
… | |
138 | our $VERSION = $AnyEvent::MP::Kernel::VERSION; |
150 | our $VERSION = $AnyEvent::MP::Kernel::VERSION; |
139 | |
151 | |
140 | our @EXPORT = qw( |
152 | our @EXPORT = qw( |
141 | NODE $NODE *SELF node_of after |
153 | NODE $NODE *SELF node_of after |
142 | configure |
154 | configure |
143 | snd rcv mon mon_guard kil reg psub spawn |
155 | snd rcv mon mon_guard kil reg psub spawn cal |
144 | port |
156 | port |
145 | ); |
157 | ); |
146 | |
158 | |
147 | our $SELF; |
159 | our $SELF; |
148 | |
160 | |
… | |
… | |
160 | |
172 | |
161 | =item $nodeid = node_of $port |
173 | =item $nodeid = node_of $port |
162 | |
174 | |
163 | Extracts and returns the node ID from a port ID or a node ID. |
175 | Extracts and returns the node ID from a port ID or a node ID. |
164 | |
176 | |
|
|
177 | =item configure $profile, key => value... |
|
|
178 | |
165 | =item configure key => value... |
179 | =item configure key => value... |
166 | |
180 | |
167 | Before a node can talk to other nodes on the network (i.e. enter |
181 | Before a node can talk to other nodes on the network (i.e. enter |
168 | "distributed mode") it has to configure itself - the minimum a node needs |
182 | "distributed mode") it has to configure itself - the minimum a node needs |
169 | to know is its own name, and optionally it should know the addresses of |
183 | to know is its own name, and optionally it should know the addresses of |
… | |
… | |
176 | |
190 | |
177 | =item step 1, gathering configuration from profiles |
191 | =item step 1, gathering configuration from profiles |
178 | |
192 | |
179 | The function first looks up a profile in the aemp configuration (see the |
193 | The function first looks up a profile in the aemp configuration (see the |
180 | L<aemp> commandline utility). The profile name can be specified via the |
194 | L<aemp> commandline utility). The profile name can be specified via the |
181 | named C<profile> parameter. If it is missing, then the nodename (F<uname |
195 | named C<profile> parameter or can simply be the first parameter). If it is |
182 | -n>) will be used as profile name. |
196 | missing, then the nodename (F<uname -n>) will be used as profile name. |
183 | |
197 | |
184 | The profile data is then gathered as follows: |
198 | The profile data is then gathered as follows: |
185 | |
199 | |
186 | First, all remaining key => value pairs (all of which are conviniently |
200 | First, all remaining key => value pairs (all of which are conveniently |
187 | undocumented at the moment) will be interpreted as configuration |
201 | undocumented at the moment) will be interpreted as configuration |
188 | data. Then they will be overwritten by any values specified in the global |
202 | data. Then they will be overwritten by any values specified in the global |
189 | default configuration (see the F<aemp> utility), then the chain of |
203 | default configuration (see the F<aemp> utility), then the chain of |
190 | profiles chosen by the profile name (and any C<parent> attributes). |
204 | profiles chosen by the profile name (and any C<parent> attributes). |
191 | |
205 | |
… | |
… | |
215 | L<AnyEvent::MP::Global> module, which will then use it to keep |
229 | L<AnyEvent::MP::Global> module, which will then use it to keep |
216 | connectivity with at least one node at any point in time. |
230 | connectivity with at least one node at any point in time. |
217 | |
231 | |
218 | =back |
232 | =back |
219 | |
233 | |
220 | Example: become a distributed node using the locla node name as profile. |
234 | Example: become a distributed node using the local node name as profile. |
221 | This should be the most common form of invocation for "daemon"-type nodes. |
235 | This should be the most common form of invocation for "daemon"-type nodes. |
222 | |
236 | |
223 | configure |
237 | configure |
224 | |
238 | |
225 | Example: become an anonymous node. This form is often used for commandline |
239 | Example: become an anonymous node. This form is often used for commandline |
… | |
… | |
474 | |
488 | |
475 | Monitor the given port and do something when the port is killed or |
489 | Monitor the given port and do something when the port is killed or |
476 | messages to it were lost, and optionally return a guard that can be used |
490 | messages to it were lost, and optionally return a guard that can be used |
477 | to stop monitoring again. |
491 | to stop monitoring again. |
478 | |
492 | |
|
|
493 | In the first form (callback), the callback is simply called with any |
|
|
494 | number of C<@reason> elements (no @reason means that the port was deleted |
|
|
495 | "normally"). Note also that I<< the callback B<must> never die >>, so use |
|
|
496 | C<eval> if unsure. |
|
|
497 | |
|
|
498 | In the second form (another port given), the other port (C<$rcvport>) |
|
|
499 | will be C<kil>'ed with C<@reason>, if a @reason was specified, i.e. on |
|
|
500 | "normal" kils nothing happens, while under all other conditions, the other |
|
|
501 | port is killed with the same reason. |
|
|
502 | |
|
|
503 | The third form (kill self) is the same as the second form, except that |
|
|
504 | C<$rvport> defaults to C<$SELF>. |
|
|
505 | |
|
|
506 | In the last form (message), a message of the form C<@msg, @reason> will be |
|
|
507 | C<snd>. |
|
|
508 | |
|
|
509 | Monitoring-actions are one-shot: once messages are lost (and a monitoring |
|
|
510 | alert was raised), they are removed and will not trigger again. |
|
|
511 | |
|
|
512 | As a rule of thumb, monitoring requests should always monitor a port from |
|
|
513 | a local port (or callback). The reason is that kill messages might get |
|
|
514 | lost, just like any other message. Another less obvious reason is that |
|
|
515 | even monitoring requests can get lost (for example, when the connection |
|
|
516 | to the other node goes down permanently). When monitoring a port locally |
|
|
517 | these problems do not exist. |
|
|
518 | |
479 | C<mon> effectively guarantees that, in the absence of hardware failures, |
519 | C<mon> effectively guarantees that, in the absence of hardware failures, |
480 | after starting the monitor, either all messages sent to the port will |
520 | after starting the monitor, either all messages sent to the port will |
481 | arrive, or the monitoring action will be invoked after possible message |
521 | arrive, or the monitoring action will be invoked after possible message |
482 | loss has been detected. No messages will be lost "in between" (after |
522 | loss has been detected. No messages will be lost "in between" (after |
483 | the first lost message no further messages will be received by the |
523 | the first lost message no further messages will be received by the |
484 | port). After the monitoring action was invoked, further messages might get |
524 | port). After the monitoring action was invoked, further messages might get |
485 | delivered again. |
525 | delivered again. |
486 | |
526 | |
487 | Note that monitoring-actions are one-shot: once messages are lost (and a |
527 | Inter-host-connection timeouts and monitoring depend on the transport |
488 | monitoring alert was raised), they are removed and will not trigger again. |
528 | used. The only transport currently implemented is TCP, and AnyEvent::MP |
|
|
529 | relies on TCP to detect node-downs (this can take 10-15 minutes on a |
|
|
530 | non-idle connection, and usually around two hours for idle conenctions). |
489 | |
531 | |
490 | In the first form (callback), the callback is simply called with any |
532 | This means that monitoring is good for program errors and cleaning up |
491 | number of C<@reason> elements (no @reason means that the port was deleted |
533 | stuff eventually, but they are no replacement for a timeout when you need |
492 | "normally"). Note also that I<< the callback B<must> never die >>, so use |
534 | to ensure some maximum latency. |
493 | C<eval> if unsure. |
|
|
494 | |
|
|
495 | In the second form (another port given), the other port (C<$rcvport>) |
|
|
496 | will be C<kil>'ed with C<@reason>, iff a @reason was specified, i.e. on |
|
|
497 | "normal" kils nothing happens, while under all other conditions, the other |
|
|
498 | port is killed with the same reason. |
|
|
499 | |
|
|
500 | The third form (kill self) is the same as the second form, except that |
|
|
501 | C<$rvport> defaults to C<$SELF>. |
|
|
502 | |
|
|
503 | In the last form (message), a message of the form C<@msg, @reason> will be |
|
|
504 | C<snd>. |
|
|
505 | |
|
|
506 | As a rule of thumb, monitoring requests should always monitor a port from |
|
|
507 | a local port (or callback). The reason is that kill messages might get |
|
|
508 | lost, just like any other message. Another less obvious reason is that |
|
|
509 | even monitoring requests can get lost (for exmaple, when the connection |
|
|
510 | to the other node goes down permanently). When monitoring a port locally |
|
|
511 | these problems do not exist. |
|
|
512 | |
535 | |
513 | Example: call a given callback when C<$port> is killed. |
536 | Example: call a given callback when C<$port> is killed. |
514 | |
537 | |
515 | mon $port, sub { warn "port died because of <@_>\n" }; |
538 | mon $port, sub { warn "port died because of <@_>\n" }; |
516 | |
539 | |
… | |
… | |
611 | the package, then the package above the package and so on (e.g. |
634 | the package, then the package above the package and so on (e.g. |
612 | C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
635 | C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
613 | exists or it runs out of package names. |
636 | exists or it runs out of package names. |
614 | |
637 | |
615 | The init function is then called with the newly-created port as context |
638 | The init function is then called with the newly-created port as context |
616 | object (C<$SELF>) and the C<@initdata> values as arguments. |
639 | object (C<$SELF>) and the C<@initdata> values as arguments. It I<must> |
|
|
640 | call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise |
|
|
641 | the port might not get created. |
617 | |
642 | |
618 | A common idiom is to pass a local port, immediately monitor the spawned |
643 | A common idiom is to pass a local port, immediately monitor the spawned |
619 | port, and in the remote init function, immediately monitor the passed |
644 | port, and in the remote init function, immediately monitor the passed |
620 | local port. This two-way monitoring ensures that both ports get cleaned up |
645 | local port. This two-way monitoring ensures that both ports get cleaned up |
621 | when there is a problem. |
646 | when there is a problem. |
622 | |
647 | |
|
|
648 | C<spawn> guarantees that the C<$initfunc> has no visible effects on the |
|
|
649 | caller before C<spawn> returns (by delaying invocation when spawn is |
|
|
650 | called for the local node). |
|
|
651 | |
623 | Example: spawn a chat server port on C<$othernode>. |
652 | Example: spawn a chat server port on C<$othernode>. |
624 | |
653 | |
625 | # this node, executed from within a port context: |
654 | # this node, executed from within a port context: |
626 | my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
655 | my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
627 | mon $server; |
656 | mon $server; |
… | |
… | |
641 | |
670 | |
642 | sub _spawn { |
671 | sub _spawn { |
643 | my $port = shift; |
672 | my $port = shift; |
644 | my $init = shift; |
673 | my $init = shift; |
645 | |
674 | |
|
|
675 | # rcv will create the actual port |
646 | local $SELF = "$NODE#$port"; |
676 | local $SELF = "$NODE#$port"; |
647 | eval { |
677 | eval { |
648 | &{ load_func $init } |
678 | &{ load_func $init } |
649 | }; |
679 | }; |
650 | _self_die if $@; |
680 | _self_die if $@; |
… | |
… | |
685 | ? $action[0]() |
715 | ? $action[0]() |
686 | : snd @action; |
716 | : snd @action; |
687 | }; |
717 | }; |
688 | } |
718 | } |
689 | |
719 | |
|
|
720 | =item cal $port, @msg, $callback[, $timeout] |
|
|
721 | |
|
|
722 | A simple form of RPC - sends a message to the given C<$port> with the |
|
|
723 | given contents (C<@msg>), but adds a reply port to the message. |
|
|
724 | |
|
|
725 | The reply port is created temporarily just for the purpose of receiving |
|
|
726 | the reply, and will be C<kil>ed when no longer needed. |
|
|
727 | |
|
|
728 | A reply message sent to the port is passed to the C<$callback> as-is. |
|
|
729 | |
|
|
730 | If an optional time-out (in seconds) is given and it is not C<undef>, |
|
|
731 | then the callback will be called without any arguments after the time-out |
|
|
732 | elapsed and the port is C<kil>ed. |
|
|
733 | |
|
|
734 | If no time-out is given, then the local port will monitor the remote port |
|
|
735 | instead, so it eventually gets cleaned-up. |
|
|
736 | |
|
|
737 | Currently this function returns the temporary port, but this "feature" |
|
|
738 | might go in future versions unless you can make a convincing case that |
|
|
739 | this is indeed useful for something. |
|
|
740 | |
|
|
741 | =cut |
|
|
742 | |
|
|
743 | sub cal(@) { |
|
|
744 | my $timeout = ref $_[-1] ? undef : pop; |
|
|
745 | my $cb = pop; |
|
|
746 | |
|
|
747 | my $port = port { |
|
|
748 | undef $timeout; |
|
|
749 | kil $SELF; |
|
|
750 | &$cb; |
|
|
751 | }; |
|
|
752 | |
|
|
753 | if (defined $timeout) { |
|
|
754 | $timeout = AE::timer $timeout, 0, sub { |
|
|
755 | undef $timeout; |
|
|
756 | kil $port; |
|
|
757 | $cb->(); |
|
|
758 | }; |
|
|
759 | } else { |
|
|
760 | mon $_[0], sub { |
|
|
761 | kil $port; |
|
|
762 | $cb->(); |
|
|
763 | }; |
|
|
764 | } |
|
|
765 | |
|
|
766 | push @_, $port; |
|
|
767 | &snd; |
|
|
768 | |
|
|
769 | $port |
|
|
770 | } |
|
|
771 | |
690 | =back |
772 | =back |
691 | |
773 | |
692 | =head1 AnyEvent::MP vs. Distributed Erlang |
774 | =head1 AnyEvent::MP vs. Distributed Erlang |
693 | |
775 | |
694 | AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
776 | AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
… | |
… | |
707 | |
789 | |
708 | =item * Node IDs are arbitrary strings in AEMP. |
790 | =item * Node IDs are arbitrary strings in AEMP. |
709 | |
791 | |
710 | Erlang relies on special naming and DNS to work everywhere in the same |
792 | Erlang relies on special naming and DNS to work everywhere in the same |
711 | way. AEMP relies on each node somehow knowing its own address(es) (e.g. by |
793 | way. AEMP relies on each node somehow knowing its own address(es) (e.g. by |
712 | configuraiton or DNS), but will otherwise discover other odes itself. |
794 | configuration or DNS), but will otherwise discover other odes itself. |
713 | |
795 | |
714 | =item * Erlang has a "remote ports are like local ports" philosophy, AEMP |
796 | =item * Erlang has a "remote ports are like local ports" philosophy, AEMP |
715 | uses "local ports are like remote ports". |
797 | uses "local ports are like remote ports". |
716 | |
798 | |
717 | The failure modes for local ports are quite different (runtime errors |
799 | The failure modes for local ports are quite different (runtime errors |
… | |
… | |
730 | |
812 | |
731 | Erlang uses processes that selectively receive messages, and therefore |
813 | Erlang uses processes that selectively receive messages, and therefore |
732 | needs a queue. AEMP is event based, queuing messages would serve no |
814 | needs a queue. AEMP is event based, queuing messages would serve no |
733 | useful purpose. For the same reason the pattern-matching abilities of |
815 | useful purpose. For the same reason the pattern-matching abilities of |
734 | AnyEvent::MP are more limited, as there is little need to be able to |
816 | AnyEvent::MP are more limited, as there is little need to be able to |
735 | filter messages without dequeing them. |
817 | filter messages without dequeuing them. |
736 | |
818 | |
737 | (But see L<Coro::MP> for a more Erlang-like process model on top of AEMP). |
819 | (But see L<Coro::MP> for a more Erlang-like process model on top of AEMP). |
738 | |
820 | |
739 | =item * Erlang sends are synchronous, AEMP sends are asynchronous. |
821 | =item * Erlang sends are synchronous, AEMP sends are asynchronous. |
740 | |
822 | |
… | |
… | |
846 | L<AnyEvent::MP::Kernel> - more, lower-level, stuff. |
928 | L<AnyEvent::MP::Kernel> - more, lower-level, stuff. |
847 | |
929 | |
848 | L<AnyEvent::MP::Global> - network maintainance and port groups, to find |
930 | L<AnyEvent::MP::Global> - network maintainance and port groups, to find |
849 | your applications. |
931 | your applications. |
850 | |
932 | |
|
|
933 | L<AnyEvent::MP::LogCatcher> - simple service to display log messages from |
|
|
934 | all nodes. |
|
|
935 | |
851 | L<AnyEvent>. |
936 | L<AnyEvent>. |
852 | |
937 | |
853 | =head1 AUTHOR |
938 | =head1 AUTHOR |
854 | |
939 | |
855 | Marc Lehmann <schmorp@schmorp.de> |
940 | Marc Lehmann <schmorp@schmorp.de> |