[ViewVC] Diff of: cvs/AnyEvent-Fork/Fork.pm

Comparing AnyEvent-Fork/Fork.pm (file contents):
Revision 1.1 by root, Sun Mar 31 03:21:27 2013 UTC vs.
Revision 1.47 by root, Thu Apr 18 20:17:34 2013 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent::ProcessPool - manage pools of perl worker processes, exec'ed or fork'ed	3	AnyEvent::Fork - everything you wanted to use fork() for, but couldn't
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use AnyEvent::ProcessPool;	7	use AnyEvent::Fork;
		8
		9	AnyEvent::Fork
		10	->new
		11	->require ("MyModule")
		12	->run ("MyModule::server", my $cv = AE::cv);
		13
		14	my $fh = $cv->recv;
8		15
9	=head1 DESCRIPTION	16	=head1 DESCRIPTION
10		17
11	This module allows you to create single worker processes but also worker	18	This module allows you to create new processes, without actually forking
12	pool that share memory, by forking from the main program, or exec'ing new	19	them from your current process (avoiding the problems of forking), but
13	perl interpreters from a module.	20	preserving most of the advantages of fork.
14		21
15	You create a new processes in a pool by specifying a function to call	22	It can be used to create new worker processes or new independent
16	with any combination of string values and file handles.	23	subprocesses for short- and long-running jobs, process pools (e.g. for use
		24	in pre-forked servers) but also to spawn new external processes (such as
		25	CGI scripts from a web server), which can be faster (and more well behaved)
		26	than using fork+exec in big processes.
17		27
18	A pool can have initialisation code which is executed before forking. The	28	Special care has been taken to make this module useful from other modules,
19	initialisation code is only executed once and the resulting process is	29	while still supporting specialised environments such as L<App::Staticperl>
20	cached, to be used as a template.	30	or L<PAR::Packer>.
21		31
22	Pools without such initialisation code don't cache an extra process.	32	=head2 WHAT THIS MODULE IS NOT
23		33
		34	This module only creates processes and lets you pass file handles and
		35	strings to it, and run perl code. It does not implement any kind of RPC -
		36	there is no back channel from the process back to you, and there is no RPC
		37	or message passing going on.
		38
		39	If you need some form of RPC, you could use the L<AnyEvent::Fork::RPC>
		40	companion module, which adds simple RPC/job queueing to a process created
		41	by this module.
		42
		43	Or you can implement it yourself in whatever way you like, use some
		44	message-passing module such as L<AnyEvent::MP>, some pipe such as
		45	L<AnyEvent::ZeroMQ>, use L<AnyEvent::Handle> on both sides to send
		46	e.g. JSON or Storable messages, and so on.
		47
		48	=head2 COMPARISON TO OTHER MODULES
		49
		50	There is an abundance of modules on CPAN that do "something fork", such as
		51	L<Parallel::ForkManager>, L<AnyEvent::ForkManager>, L<AnyEvent::Worker>
		52	or L<AnyEvent::Subprocess>. There are modules that implement their own
		53	process management, such as L<AnyEvent::DBI>.
		54
		55	The problems that all these modules try to solve are real, however, none
		56	of them (from what I have seen) tackle the very real problems of unwanted
		57	memory sharing, efficiency, not being able to use event processing or
		58	similar modules in the processes they create.
		59
		60	This module doesn't try to replace any of them - instead it tries to solve
		61	the problem of creating processes with a minimum of fuss and overhead (and
		62	also luxury). Ideally, most of these would use AnyEvent::Fork internally,
		63	except they were written before AnyEvent:Fork was available, so obviously
		64	had to roll their own.
		65
24	=head1 PROBLEM STATEMENT	66	=head2 PROBLEM STATEMENT
25		67
26	There are two ways to implement parallel processing on UNIX like operating	68	There are two traditional ways to implement parallel processing on UNIX
27	systems - fork and process, and fork+exec and process. They have different	69	like operating systems - fork and process, and fork+exec and process. They
28	advantages and disadvantages that I describe below, together with how this	70	have different advantages and disadvantages that I describe below,
29	module tries to mitigate the disadvantages.	71	together with how this module tries to mitigate the disadvantages.
30		72
31	=over 4	73	=over 4
32		74
33	=item Forking from a big process can be very slow (a 5GB process needs	75	=item Forking from a big process can be very slow.
34	0.05s to fork on my 3.6GHz amd64 GNU/Linux box for example). This overhead	76
		77	A 5GB process needs 0.05s to fork on my 3.6GHz amd64 GNU/Linux box. This
35	is often shared with exec (because you have to fork first), but in some	78	overhead is often shared with exec (because you have to fork first), but
36	circumstances (e.g. when vfork is used), fork+exec can be much faster.	79	in some circumstances (e.g. when vfork is used), fork+exec can be much
		80	faster.
37		81
38	This module can help here by telling a small(er) helper process to fork,	82	This module can help here by telling a small(er) helper process to fork,
39	or fork+exec instead.	83	which is faster then forking the main process, and also uses vfork where
		84	possible. This gives the speed of vfork, with the flexibility of fork.
40		85
41	=item Forking usually creates a copy-on-write copy of the parent	86	=item Forking usually creates a copy-on-write copy of the parent
42	process. Memory (for example, modules or data files that have been	87	process.
43	will not take additional memory). When exec'ing a new process, modules	88
44	and data files might need to be loaded again, at extra cpu and memory	89	For example, modules or data files that are loaded will not use additional
45	cost. Likewise when forking, all data structures are copied as well - if	90	memory after a fork. When exec'ing a new process, modules and data files
		91	might need to be loaded again, at extra CPU and memory cost. But when
		92	forking, literally all data structures are copied - if the program frees
46	the program frees them and replaces them by new data, the child processes	93	them and replaces them by new data, the child processes will retain the
47	will retain the memory even if it isn't used.	94	old version even if it isn't used, which can suddenly and unexpectedly
		95	increase memory usage when freeing memory.
		96
		97	The trade-off is between more sharing with fork (which can be good or
		98	bad), and no sharing with exec.
48		99
49	This module allows the main program to do a controlled fork, and allows	100	This module allows the main program to do a controlled fork, and allows
50	modules to exec processes safely at any time. When creating a custom	101	modules to exec processes safely at any time. When creating a custom
51	process pool you can take advantage of data sharing via fork without	102	process pool you can take advantage of data sharing via fork without
52	risking to share large dynamic data structures that will blow up child	103	risking to share large dynamic data structures that will blow up child
53	memory usage.	104	memory usage.
54		105
		106	In other words, this module puts you into control over what is being
		107	shared and what isn't, at all times.
		108
55	=item Exec'ing a new perl process might be difficult and slow. For	109	=item Exec'ing a new perl process might be difficult.
		110
56	example, it is not easy to find the correct path to the perl interpreter,	111	For example, it is not easy to find the correct path to the perl
57	and all modules have to be loaded from disk again. Long running processes	112	interpreter - C<$^X> might not be a perl interpreter at all.
58	might run into problems when perl is upgraded for example.
59		113
60	This module supports creating pre-initialised perl processes to be used
61	as template, and also tries hard to identify the correct path to the perl	114	This module tries hard to identify the correct path to the perl
62	interpreter. With a cooperative main program, exec'ing the interpreter	115	interpreter. With a cooperative main program, exec'ing the interpreter
63	might not even be necessary.	116	might not even be necessary, but even without help from the main program,
		117	it will still work when used from a module.
64		118
		119	=item Exec'ing a new perl process might be slow, as all necessary modules
		120	have to be loaded from disk again, with no guarantees of success.
		121
		122	Long running processes might run into problems when perl is upgraded
		123	and modules are no longer loadable because they refer to a different
		124	perl version, or parts of a distribution are newer than the ones already
		125	loaded.
		126
		127	This module supports creating pre-initialised perl processes to be used as
		128	a template for new processes.
		129
65	=item Forking might be impossible when a program is running. For example,	130	=item Forking might be impossible when a program is running.
66	POSIX makes it almost impossible to fork from a multithreaded program and
67	do anything useful in the child - strictly speaking, if your perl program
68	uses posix threads (even indirectly via e.g. L<IO::AIO> or L<threads>),
69	you cannot call fork on the perl level anymore, at all.
70		131
		132	For example, POSIX makes it almost impossible to fork from a
		133	multi-threaded program while doing anything useful in the child - in
		134	fact, if your perl program uses POSIX threads (even indirectly via
		135	e.g. L<IO::AIO> or L<threads>), you cannot call fork on the perl level
		136	anymore without risking corruption issues on a number of operating
		137	systems.
		138
71	This module can safely fork helper processes at any time, by caling	139	This module can safely fork helper processes at any time, by calling
72	fork+exec in C, in a POSIX-compatible way.	140	fork+exec in C, in a POSIX-compatible way (via L<Proc::FastSpawn>).
73		141
74	=item Parallel processing with fork might be inconvenient or difficult	142	=item Parallel processing with fork might be inconvenient or difficult
		143	to implement. Modules might not work in both parent and child.
		144
75	to implement. For example, when a program uses an event loop and creates	145	For example, when a program uses an event loop and creates watchers it
76	watchers it becomes very hard to use the event loop from a child	146	becomes very hard to use the event loop from a child program, as the
77	program, as the watchers already exist but are only meaningful in the	147	watchers already exist but are only meaningful in the parent. Worse, a
78	parent. Worse, a module might want to use such a system, not knowing	148	module might want to use such a module, not knowing whether another module
79	whether another module or the main program also does, leading to problems.	149	or the main program also does, leading to problems.
80		150
81	This module only lets the main program create pools by forking (because	151	Apart from event loops, graphical toolkits also commonly fall into the
82	only the main program can know when it is still safe to do so) - all other	152	"unsafe module" category, or just about anything that communicates with
83	pools are created by fork+exec, after which such modules can again be	153	the external world, such as network libraries and file I/O modules, which
84	loaded.	154	usually don't like being copied and then allowed to continue in two
		155	processes.
		156
		157	With this module only the main program is allowed to create new processes
		158	by forking (because only the main program can know when it is still safe
		159	to do so) - all other processes are created via fork+exec, which makes it
		160	possible to use modules such as event loops or window interfaces safely.
85		161
86	=back	162	=back
87		163
		164	=head1 EXAMPLES
		165
		166	=head2 Create a single new process, tell it to run your worker function.
		167
		168	AnyEvent::Fork
		169	->new
		170	->require ("MyModule")
		171	->run ("MyModule::worker, sub {
		172	my ($master_filehandle) = @_;
		173
		174	# now $master_filehandle is connected to the
		175	# $slave_filehandle in the new process.
		176	});
		177
		178	C<MyModule> might look like this:
		179
		180	package MyModule;
		181
		182	sub worker {
		183	my ($slave_filehandle) = @_;
		184
		185	# now $slave_filehandle is connected to the $master_filehandle
		186	# in the original prorcess. have fun!
		187	}
		188
		189	=head2 Create a pool of server processes all accepting on the same socket.
		190
		191	# create listener socket
		192	my $listener = ...;
		193
		194	# create a pool template, initialise it and give it the socket
		195	my $pool = AnyEvent::Fork
		196	->new
		197	->require ("Some::Stuff", "My::Server")
		198	->send_fh ($listener);
		199
		200	# now create 10 identical workers
		201	for my $id (1..10) {
		202	$pool
		203	->fork
		204	->send_arg ($id)
		205	->run ("My::Server::run");
		206	}
		207
		208	# now do other things - maybe use the filehandle provided by run
		209	# to wait for the processes to die. or whatever.
		210
		211	C<My::Server> might look like this:
		212
		213	package My::Server;
		214
		215	sub run {
		216	my ($slave, $listener, $id) = @_;
		217
		218	close $slave; # we do not use the socket, so close it to save resources
		219
		220	# we could go ballistic and use e.g. AnyEvent here, or IO::AIO,
		221	# or anything we usually couldn't do in a process forked normally.
		222	while (my $socket = $listener->accept) {
		223	# do sth. with new socket
		224	}
		225	}
		226
		227	=head2 use AnyEvent::Fork as a faster fork+exec
		228
		229	This runs C</bin/echo hi>, with standard output redirected to F</tmp/log>
		230	and standard error redirected to the communications socket. It is usually
		231	faster than fork+exec, but still lets you prepare the environment.
		232
		233	open my $output, ">/tmp/log" or die "$!";
		234
		235	AnyEvent::Fork
		236	->new
		237	->eval ('
		238	# compile a helper function for later use
		239	sub run {
		240	my ($fh, $output, @cmd) = @_;
		241
		242	# perl will clear close-on-exec on STDOUT/STDERR
		243	open STDOUT, ">&", $output or die;
		244	open STDERR, ">&", $fh or die;
		245
		246	exec @cmd;
		247	}
		248	')
		249	->send_fh ($output)
		250	->send_arg ("/bin/echo", "hi")
		251	->run ("run", my $cv = AE::cv);
		252
		253	my $stderr = $cv->recv;
		254
		255	=head1 CONCEPTS
		256
		257	This module can create new processes either by executing a new perl
		258	process, or by forking from an existing "template" process.
		259
		260	All these processes are called "child processes" (whether they are direct
		261	children or not), while the process that manages them is called the
		262	"parent process".
		263
		264	Each such process comes with its own file handle that can be used to
		265	communicate with it (it's actually a socket - one end in the new process,
		266	one end in the main process), and among the things you can do in it are
		267	load modules, fork new processes, send file handles to it, and execute
		268	functions.
		269
		270	There are multiple ways to create additional processes to execute some
		271	jobs:
		272
88	=over 4	273	=over 4
89		274
90	=cut	275	=item fork a new process from the "default" template process, load code,
		276	run it
91		277
		278	This module has a "default" template process which it executes when it is
		279	needed the first time. Forking from this process shares the memory used
		280	for the perl interpreter with the new process, but loading modules takes
		281	time, and the memory is not shared with anything else.
		282
		283	This is ideal for when you only need one extra process of a kind, with the
		284	option of starting and stopping it on demand.
		285
		286	Example:
		287
		288	AnyEvent::Fork
		289	->new
		290	->require ("Some::Module")
		291	->run ("Some::Module::run", sub {
		292	my ($fork_fh) = @_;
		293	});
		294
		295	=item fork a new template process, load code, then fork processes off of
		296	it and run the code
		297
		298	When you need to have a bunch of processes that all execute the same (or
		299	very similar) tasks, then a good way is to create a new template process
		300	for them, loading all the modules you need, and then create your worker
		301	processes from this new template process.
		302
		303	This way, all code (and data structures) that can be shared (e.g. the
		304	modules you loaded) is shared between the processes, and each new process
		305	consumes relatively little memory of its own.
		306
		307	The disadvantage of this approach is that you need to create a template
		308	process for the sole purpose of forking new processes from it, but if you
		309	only need a fixed number of processes you can create them, and then destroy
		310	the template process.
		311
		312	Example:
		313
		314	my $template = AnyEvent::Fork->new->require ("Some::Module");
		315
		316	for (1..10) {
		317	$template->fork->run ("Some::Module::run", sub {
		318	my ($fork_fh) = @_;
		319	});
		320	}
		321
		322	# at this point, you can keep $template around to fork new processes
		323	# later, or you can destroy it, which causes it to vanish.
		324
		325	=item execute a new perl interpreter, load some code, run it
		326
		327	This is relatively slow, and doesn't allow you to share memory between
		328	multiple processes.
		329
		330	The only advantage is that you don't have to have a template process
		331	hanging around all the time to fork off some new processes, which might be
		332	an advantage when there are long time spans where no extra processes are
		333	needed.
		334
		335	Example:
		336
		337	AnyEvent::Fork
		338	->new_exec
		339	->require ("Some::Module")
		340	->run ("Some::Module::run", sub {
		341	my ($fork_fh) = @_;
		342	});
		343
		344	=back
		345
		346	=head1 THE C<AnyEvent::Fork> CLASS
		347
		348	This module exports nothing, and only implements a single class -
		349	C<AnyEvent::Fork>.
		350
		351	There are two class constructors that both create new processes - C<new>
		352	and C<new_exec>. The C<fork> method creates a new process by forking an
		353	existing one and could be considered a third constructor.
		354
		355	Most of the remaining methods deal with preparing the new process, by
		356	loading code, evaluating code and sending data to the new process. They
		357	usually return the process object, so you can chain method calls.
		358
		359	If a process object is destroyed before calling its C<run> method, then
		360	the process simply exits. After C<run> is called, all responsibility is
		361	passed to the specified function.
		362
		363	As long as there is any outstanding work to be done, process objects
		364	resist being destroyed, so there is no reason to store them unless you
		365	need them later - configure and forget works just fine.
		366
		367	=over 4
		368
		369	=cut
		370
92	package AnyEvent::ProcessPool;	371	package AnyEvent::Fork;
93		372
94	use common::sense;	373	use common::sense;
95		374
96	use Socket ();	375	use Errno ();
97		376
98	use Proc::FastSpawn;
99	use AnyEvent;	377	use AnyEvent;
100	use AnyEvent::ProcessPool::Util;
101	use AnyEvent::Util ();	378	use AnyEvent::Util ();
102		379
103	BEGIN {	380	use IO::FDPass;
104	# require Exporter;
105	}
106		381
107	=item my $pool = new AnyEvent::ProcessPool key => value...	382	our $VERSION = 0.7;
108		383
109	Create a new process pool. The following named parameters are supported:	384	# the early fork template process
		385	our $EARLY;
110		386
111	=over 4
112
113	=back
114
115	=cut
116
117	# the template process	387	# the empty template process
118	our $template;	388	our $TEMPLATE;
119		389
120	sub _queue {	390	sub QUEUE() { 0 }
121	my ($pid, $fh) = @_;	391	sub FH() { 1 }
		392	sub WW() { 2 }
		393	sub PID() { 3 }
		394	sub CB() { 4 }
122		395
123	[	396	sub _new {
		397	my ($self, $fh, $pid) = @_;
		398
		399	AnyEvent::Util::fh_nonblocking $fh, 1;
		400
		401	$self = bless [
		402	[], # write queue - strings or fd's
		403	$fh,
		404	undef, # AE watcher
124	$pid,	405	$pid,
125	$fh,	406	], $self;
126	[],
127	undef
128	]
129	}
130		407
		408	$self
		409	}
		410
131	sub queue_cmd {	411	sub _cmd {
132	my ($queue, $cmd) = @_;	412	my $self = shift;
133		413
134	push @{ $queue->[2] }, pack "N/a", $cmd;	414	# ideally, we would want to use "a (w/a)*" as format string, but perl
		415	# versions from at least 5.8.9 to 5.16.3 are all buggy and can't unpack
		416	# it.
		417	push @{ $self->[QUEUE] }, pack "a L/a*", $_[0], $_[1];
135		418
136	$queue->[3] \|\|= AE::io $queue->[1], 1, sub {	419	$self->[WW] \|\|= AE::io $self->[FH], 1, sub {
137	warn "oopl0 ", scalar @{ $queue->[2] };	420	do {
		421	# send the next "thing" in the queue - either a reference to an fh,
		422	# or a plain string.
		423
138	if (ref $queue->[2][0]) {	424	if (ref $self->[QUEUE][0]) {
139	warn "oopla2\n";#d#	425	# send fh
140	AnyEvent::ProcessPool::Util::fd_send fileno $queue->[1], fileno ${ $queue->[2][0] }	426	unless (IO::FDPass::send fileno $self->[FH], fileno ${ $self->[QUEUE][0] }) {
		427	return if $! == Errno::EAGAIN \|\| $! == Errno::EWOULDBLOCK;
		428	undef $self->[WW];
		429	die "AnyEvent::Fork: file descriptor send failure: $!";
		430	}
		431
141	and shift @{ $queue->[2] };	432	shift @{ $self->[QUEUE] };
		433
142	} else {	434	} else {
143	warn "write ", length $queue->[2][0];#d#	435	# send string
144	my $len = syswrite $queue->[1], $queue->[2][0]	436	my $len = syswrite $self->[FH], $self->[QUEUE][0];
		437
		438	unless ($len) {
		439	return if $! == Errno::EAGAIN \|\| $! == Errno::EWOULDBLOCK;
		440	undef $self->[3];
145	or die "AnyEvent::ProcessPool::queue write failure: $!";	441	die "AnyEvent::Fork: command write failure: $!";
		442	}
		443
146	substr $queue->[2][0], 0, $len, "";	444	substr $self->[QUEUE][0], 0, $len, "";
147	shift @{ $queue->[2] } unless length $queue->[2][0];	445	shift @{ $self->[QUEUE] } unless length $self->[QUEUE][0];
148	}	446	}
		447	} while @{ $self->[QUEUE] };
149		448
150	undef $queue->[3] unless @{ $queue->[2] };	449	# everything written
151	warn "oopl3 ", scalar @{ $queue->[2] };	450	undef $self->[WW];
152	warn "oopl4 $queue->[3]\n";#d#	451
		452	# invoke run callback, if any
		453	$self->[CB]->($self->[FH]) if $self->[CB];
153	};	454	};
154	}
155		455
156	sub run_template {	456	() # make sure we don't leak the watcher
157	return if $template;	457	}
158		458
		459	# fork template from current process, used by AnyEvent::Fork::Early/Template
		460	sub _new_fork {
159	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;	461	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;
160	AnyEvent::Util::fh_nonblocking $fh, 1;	462	my $parent = $$;
161	fd_inherit fileno $slave;
162		463
163	my %env = %ENV;	464	my $pid = fork;
164	$env{PERL5LIB} = join ":", grep !ref, @INC;
165		465
166	my $pid = spawn	466	if ($pid eq 0) {
167	$^X,	467	require AnyEvent::Fork::Serve;
168	["perl", "-MAnyEvent::ProcessPool::Serve", "-e", "AnyEvent::ProcessPool::Serve::me", fileno $slave],	468	$AnyEvent::Fork::Serve::OWNER = $parent;
169	[map "$_=$env{$_}", keys %env],	469	close $fh;
170	or die "unable to spawn AnyEvent::ProcessPool server: $!";	470	$0 = "$_[1] of $parent";
		471	AnyEvent::Fork::Serve::serve ($slave);
		472	exit 0;
		473	} elsif (!$pid) {
		474	die "AnyEvent::Fork::Early/Template: unable to fork template process: $!";
		475	}
171		476
172	close $slave;	477	AnyEvent::Fork->_new ($fh, $pid)
173
174	$template = _queue $pid, $fh;
175
176	my ($a, $b) = AnyEvent::Util::portable_socketpair;
177
178	queue_cmd $template, "Iabc";
179	push @{ $template->[2] }, \$b;
180
181	use Coro::AnyEvent; Coro::AnyEvent::sleep 1;
182	}	478	}
		479
		480	=item my $proc = new AnyEvent::Fork
		481
		482	Create a new "empty" perl interpreter process and returns its process
		483	object for further manipulation.
		484
		485	The new process is forked from a template process that is kept around
		486	for this purpose. When it doesn't exist yet, it is created by a call to
		487	C<new_exec> first and then stays around for future calls.
		488
		489	=cut
183		490
184	sub new {	491	sub new {
185	my $class = shift;	492	my $class = shift;
186		493
187	my $self = bless {	494	$TEMPLATE \|\|= $class->new_exec;
188	@_	495	$TEMPLATE->fork
189	}, $class;	496	}
190		497
191	run_template;	498	=item $new_proc = $proc->fork
		499
		500	Forks C<$proc>, creating a new process, and returns the process object
		501	of the new process.
		502
		503	If any of the C<send_> functions have been called before fork, then they
		504	will be cloned in the child. For example, in a pre-forked server, you
		505	might C<send_fh> the listening socket into the template process, and then
		506	keep calling C<fork> and C<run>.
		507
		508	=cut
		509
		510	sub fork {
		511	my ($self) = @_;
		512
		513	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;
		514
		515	$self->send_fh ($slave);
		516	$self->_cmd ("f");
		517
		518	AnyEvent::Fork->_new ($fh)
		519	}
		520
		521	=item my $proc = new_exec AnyEvent::Fork
		522
		523	Create a new "empty" perl interpreter process and returns its process
		524	object for further manipulation.
		525
		526	Unlike the C<new> method, this method I<always> spawns a new perl process
		527	(except in some cases, see L<AnyEvent::Fork::Early> for details). This
		528	reduces the amount of memory sharing that is possible, and is also slower.
		529
		530	You should use C<new> whenever possible, except when having a template
		531	process around is unacceptable.
		532
		533	The path to the perl interpreter is divined using various methods - first
		534	C<$^X> is investigated to see if the path ends with something that sounds
		535	as if it were the perl interpreter. Failing this, the module falls back to
		536	using C<$Config::Config{perlpath}>.
		537
		538	=cut
		539
		540	sub new_exec {
		541	my ($self) = @_;
		542
		543	return $EARLY->fork
		544	if $EARLY;
		545
		546	# first find path of perl
		547	my $perl = $;
		548
		549	# first we try $^X, but the path must be absolute (always on win32), and end in sth.
		550	# that looks like perl. this obviously only works for posix and win32
		551	unless (
		552	($^O eq "MSWin32" \|\| $perl =~ m%^/%)
		553	&& $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i
		554	) {
		555	# if it doesn't look perlish enough, try Config
		556	require Config;
		557	$perl = $Config::Config{perlpath};
		558	$perl =~ s/(?:\Q$Config::Config{_exe}\E)?$/$Config::Config{_exe}/;
		559	}
		560
		561	require Proc::FastSpawn;
		562
		563	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;
		564	Proc::FastSpawn::fd_inherit (fileno $slave);
		565
		566	# new fh's should always be set cloexec (due to $^F),
		567	# but hey, not on win32, so we always clear the inherit flag.
		568	Proc::FastSpawn::fd_inherit (fileno $fh, 0);
		569
		570	# quick. also doesn't work in win32. of course. what did you expect
		571	#local $ENV{PERL5LIB} = join ":", grep !ref, @INC;
		572	my %env = %ENV;
		573	$env{PERL5LIB} = join +($^O eq "MSWin32" ? ";" : ":"), grep !ref, @INC;
		574
		575	my $pid = Proc::FastSpawn::spawn (
		576	$perl,
		577	["perl", "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$],
		578	[map "$_=$env{$_}", keys %env],
		579	) or die "unable to spawn AnyEvent::Fork server: $!";
		580
		581	$self->_new ($fh, $pid)
		582	}
		583
		584	=item $pid = $proc->pid
		585
		586	Returns the process id of the process I<iff it is a direct child of the
		587	process running AnyEvent::Fork>, and C<undef> otherwise.
		588
		589	Normally, only processes created via C<< AnyEvent::Fork->new_exec >> and
		590	L<AnyEvent::Fork::Template> are direct children, and you are responsible
		591	to clean up their zombies when they die.
		592
		593	All other processes are not direct children, and will be cleaned up by
		594	AnyEvent::Fork itself.
		595
		596	=cut
		597
		598	sub pid {
		599	$_[0][PID]
		600	}
		601
		602	=item $proc = $proc->eval ($perlcode, @args)
		603
		604	Evaluates the given C<$perlcode> as ... Perl code, while setting C<@_> to
		605	the strings specified by C<@args>, in the "main" package.
		606
		607	This call is meant to do any custom initialisation that might be required
		608	(for example, the C<require> method uses it). It's not supposed to be used
		609	to completely take over the process, use C<run> for that.
		610
		611	The code will usually be executed after this call returns, and there is no
		612	way to pass anything back to the calling process. Any evaluation errors
		613	will be reported to stderr and cause the process to exit.
		614
		615	If you want to execute some code (that isn't in a module) to take over the
		616	process, you should compile a function via C<eval> first, and then call
		617	it via C<run>. This also gives you access to any arguments passed via the
		618	C<send_xxx> methods, such as file handles. See the L<use AnyEvent::Fork as
		619	a faster fork+exec> example to see it in action.
		620
		621	Returns the process object for easy chaining of method calls.
		622
		623	=cut
		624
		625	sub eval {
		626	my ($self, $code, @args) = @_;
		627
		628	$self->_cmd (e => pack "(w/a)", $code, @args);
192		629
193	$self	630	$self
194	}	631	}
195		632
		633	=item $proc = $proc->require ($module, ...)
		634
		635	Tries to load the given module(s) into the process
		636
		637	Returns the process object for easy chaining of method calls.
		638
		639	=cut
		640
		641	sub require {
		642	my ($self, @modules) = @_;
		643
		644	s%::%/%g for @modules;
		645	$self->eval ('require "$_.pm" for @_', @modules);
		646
		647	$self
		648	}
		649
		650	=item $proc = $proc->send_fh ($handle, ...)
		651
		652	Send one or more file handles (I<not> file descriptors) to the process,
		653	to prepare a call to C<run>.
		654
		655	The process object keeps a reference to the handles until they have
		656	been passed over to the process, so you must not explicitly close the
		657	handles. This is most easily accomplished by simply not storing the file
		658	handles anywhere after passing them to this method - when AnyEvent::Fork
		659	is finished using them, perl will automatically close them.
		660
		661	Returns the process object for easy chaining of method calls.
		662
		663	Example: pass a file handle to a process, and release it without
		664	closing. It will be closed automatically when it is no longer used.
		665
		666	$proc->send_fh ($my_fh);
		667	undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT
		668
		669	=cut
		670
		671	sub send_fh {
		672	my ($self, @fh) = @_;
		673
		674	for my $fh (@fh) {
		675	$self->_cmd ("h");
		676	push @{ $self->[QUEUE] }, \$fh;
		677	}
		678
		679	$self
		680	}
		681
		682	=item $proc = $proc->send_arg ($string, ...)
		683
		684	Send one or more argument strings to the process, to prepare a call to
		685	C<run>. The strings can be any octet strings.
		686
		687	The protocol is optimised to pass a moderate number of relatively short
		688	strings - while you can pass up to 4GB of data in one go, this is more
		689	meant to pass some ID information or other startup info, not big chunks of
		690	data.
		691
		692	Returns the process object for easy chaining of method calls.
		693
		694	=cut
		695
		696	sub send_arg {
		697	my ($self, @arg) = @_;
		698
		699	$self->_cmd (a => pack "(w/a)", @arg);
		700
		701	$self
		702	}
		703
		704	=item $proc->run ($func, $cb->($fh))
		705
		706	Enter the function specified by the function name in C<$func> in the
		707	process. The function is called with the communication socket as first
		708	argument, followed by all file handles and string arguments sent earlier
		709	via C<send_fh> and C<send_arg> methods, in the order they were called.
		710
		711	The process object becomes unusable on return from this function - any
		712	further method calls result in undefined behaviour.
		713
		714	The function name should be fully qualified, but if it isn't, it will be
		715	looked up in the C<main> package.
		716
		717	If the called function returns, doesn't exist, or any error occurs, the
		718	process exits.
		719
		720	Preparing the process is done in the background - when all commands have
		721	been sent, the callback is invoked with the local communications socket
		722	as argument. At this point you can start using the socket in any way you
		723	like.
		724
		725	If the communication socket isn't used, it should be closed on both sides,
		726	to save on kernel memory.
		727
		728	The socket is non-blocking in the parent, and blocking in the newly
		729	created process. The close-on-exec flag is set in both.
		730
		731	Even if not used otherwise, the socket can be a good indicator for the
		732	existence of the process - if the other process exits, you get a readable
		733	event on it, because exiting the process closes the socket (if it didn't
		734	create any children using fork).
		735
		736	Example: create a template for a process pool, pass a few strings, some
		737	file handles, then fork, pass one more string, and run some code.
		738
		739	my $pool = AnyEvent::Fork
		740	->new
		741	->send_arg ("str1", "str2")
		742	->send_fh ($fh1, $fh2);
		743
		744	for (1..2) {
		745	$pool
		746	->fork
		747	->send_arg ("str3")
		748	->run ("Some::function", sub {
		749	my ($fh) = @_;
		750
		751	# fh is nonblocking, but we trust that the OS can accept these
		752	# few octets anyway.
		753	syswrite $fh, "hi #$_\n";
		754
		755	# $fh is being closed here, as we don't store it anywhere
		756	});
		757	}
		758
		759	# Some::function might look like this - all parameters passed before fork
		760	# and after will be passed, in order, after the communications socket.
		761	sub Some::function {
		762	my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_;
		763
		764	print scalar <$fh>; # prints "hi #1\n" and "hi #2\n" in any order
		765	}
		766
		767	=cut
		768
		769	sub run {
		770	my ($self, $func, $cb) = @_;
		771
		772	$self->[CB] = $cb;
		773	$self->_cmd (r => $func);
		774	}
		775
196	=back	776	=back
197		777
198	=head1 AUTHOR	778	=head1 PERFORMANCE
		779
		780	Now for some unscientific benchmark numbers (all done on an amd64
		781	GNU/Linux box). These are intended to give you an idea of the relative
		782	performance you can expect, they are not meant to be absolute performance
		783	numbers.
		784
		785	OK, so, I ran a simple benchmark that creates a socket pair, forks, calls
		786	exit in the child and waits for the socket to close in the parent. I did
		787	load AnyEvent, EV and AnyEvent::Fork, for a total process size of 5100kB.
		788
		789	2079 new processes per second, using manual socketpair + fork
		790
		791	Then I did the same thing, but instead of calling fork, I called
		792	AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the
		793	socket form the child to close on exit. This does the same thing as manual
		794	socket pair + fork, except that what is forked is the template process
		795	(2440kB), and the socket needs to be passed to the server at the other end
		796	of the socket first.
		797
		798	2307 new processes per second, using AnyEvent::Fork->new
		799
		800	And finally, using C<new_exec> instead C<new>, using vforks+execs to exec
		801	a new perl interpreter and compile the small server each time, I get:
		802
		803	479 vfork+execs per second, using AnyEvent::Fork->new_exec
		804
		805	So how can C<< AnyEvent->new >> be faster than a standard fork, even
		806	though it uses the same operations, but adds a lot of overhead?
		807
		808	The difference is simply the process size: forking the 5MB process takes
		809	so much longer than forking the 2.5MB template process that the extra
		810	overhead is canceled out.
		811
		812	If the benchmark process grows, the normal fork becomes even slower:
		813
		814	1340 new processes, manual fork of a 20MB process
		815	731 new processes, manual fork of a 200MB process
		816	235 new processes, manual fork of a 2000MB process
		817
		818	What that means (to me) is that I can use this module without having a bad
		819	conscience because of the extra overhead required to start new processes.
		820
		821	=head1 TYPICAL PROBLEMS
		822
		823	This section lists typical problems that remain. I hope by recognising
		824	them, most can be avoided.
		825
		826	=over 4
		827
		828	=item leaked file descriptors for exec'ed processes
		829
		830	POSIX systems inherit file descriptors by default when exec'ing a new
		831	process. While perl itself laudably sets the close-on-exec flags on new
		832	file handles, most C libraries don't care, and even if all cared, it's
		833	often not possible to set the flag in a race-free manner.
		834
		835	That means some file descriptors can leak through. And since it isn't
		836	possible to know which file descriptors are "good" and "necessary" (or
		837	even to know which file descriptors are open), there is no good way to
		838	close the ones that might harm.
		839
		840	As an example of what "harm" can be done consider a web server that
		841	accepts connections and afterwards some module uses AnyEvent::Fork for the
		842	first time, causing it to fork and exec a new process, which might inherit
		843	the network socket. When the server closes the socket, it is still open
		844	in the child (which doesn't even know that) and the client might conclude
		845	that the connection is still fine.
		846
		847	For the main program, there are multiple remedies available -
		848	L<AnyEvent::Fork::Early> is one, creating a process early and not using
		849	C<new_exec> is another, as in both cases, the first process can be exec'ed
		850	well before many random file descriptors are open.
		851
		852	In general, the solution for these kind of problems is to fix the
		853	libraries or the code that leaks those file descriptors.
		854
		855	Fortunately, most of these leaked descriptors do no harm, other than
		856	sitting on some resources.
		857
		858	=item leaked file descriptors for fork'ed processes
		859
		860	Normally, L<AnyEvent::Fork> does start new processes by exec'ing them,
		861	which closes file descriptors not marked for being inherited.
		862
		863	However, L<AnyEvent::Fork::Early> and L<AnyEvent::Fork::Template> offer
		864	a way to create these processes by forking, and this leaks more file
		865	descriptors than exec'ing them, as there is no way to mark descriptors as
		866	"close on fork".
		867
		868	An example would be modules like L<EV>, L<IO::AIO> or L<Gtk2>. Both create
		869	pipes for internal uses, and L<Gtk2> might open a connection to the X
		870	server. L<EV> and L<IO::AIO> can deal with fork, but Gtk2 might have
		871	trouble with a fork.
		872
		873	The solution is to either not load these modules before use'ing
		874	L<AnyEvent::Fork::Early> or L<AnyEvent::Fork::Template>, or to delay
		875	initialising them, for example, by calling C<init Gtk2> manually.
		876
		877	=item exiting calls object destructors
		878
		879	This only applies to users of L<AnyEvent::Fork:Early> and
		880	L<AnyEvent::Fork::Template>, or when initialising code creates objects
		881	that reference external resources.
		882
		883	When a process created by AnyEvent::Fork exits, it might do so by calling
		884	exit, or simply letting perl reach the end of the program. At which point
		885	Perl runs all destructors.
		886
		887	Not all destructors are fork-safe - for example, an object that represents
		888	the connection to an X display might tell the X server to free resources,
		889	which is inconvenient when the "real" object in the parent still needs to
		890	use them.
		891
		892	This is obviously not a problem for L<AnyEvent::Fork::Early>, as you used
		893	it as the very first thing, right?
		894
		895	It is a problem for L<AnyEvent::Fork::Template> though - and the solution
		896	is to not create objects with nontrivial destructors that might have an
		897	effect outside of Perl.
		898
		899	=back
		900
		901	=head1 PORTABILITY NOTES
		902
		903	Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop,
		904	and ::Template is not going to work), and it cost a lot of blood and sweat
		905	to make it so, mostly due to the bloody broken perl that nobody seems to
		906	care about. The fork emulation is a bad joke - I have yet to see something
		907	useful that you can do with it without running into memory corruption
		908	issues or other braindamage. Hrrrr.
		909
		910	Cygwin perl is not supported at the moment due to some hilarious
		911	shortcomings of its API - see L<IO::FDPoll> for more details.
		912
		913	=head1 SEE ALSO
		914
		915	L<AnyEvent::Fork::Early>, to avoid executing a perl interpreter at all
		916	(part of this distribution).
		917
		918	L<AnyEvent::Fork::Template>, to create a process by forking the main
		919	program at a convenient time (part of this distribution).
		920
		921	L<AnyEvent::Fork::RPC>, for simple RPC to child processes (on CPAN).
		922
		923	=head1 AUTHOR AND CONTACT INFORMATION
199		924
200	Marc Lehmann <schmorp@schmorp.de>	925	Marc Lehmann <schmorp@schmorp.de>
201	http://home.schmorp.de/	926	http://software.schmorp.de/pkg/AnyEvent-Fork
202		927
203	=cut	928	=cut
204		929
205	1	930	1
206		931

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent-Fork/Fork.pm (file contents): Revision 1.1 by root, Sun Mar 31 03:21:27 2013 UTC vs. Revision 1.47 by root, Thu Apr 18 20:17:34 2013 UTC

Diff Legend

Comparing AnyEvent-Fork/Fork.pm (file contents):
Revision 1.1 by root, Sun Mar 31 03:21:27 2013 UTC vs.
Revision 1.47 by root, Thu Apr 18 20:17:34 2013 UTC