1 | =head1 NAME |
1 | =head1 NAME |
2 | |
2 | |
3 | AnyEvent::Fork - everything you wanted to use fork() for, but couldn't |
3 | AnyEvent::Fork - everything you wanted to use fork() for, but couldn't |
4 | |
|
|
5 | ATTENTION, this is a very early release, and very untested. Consider it a |
|
|
6 | technology preview. |
|
|
7 | |
4 | |
8 | =head1 SYNOPSIS |
5 | =head1 SYNOPSIS |
9 | |
6 | |
10 | use AnyEvent::Fork; |
7 | use AnyEvent::Fork; |
11 | |
8 | |
… | |
… | |
79 | than using fork+exec in big processes. |
76 | than using fork+exec in big processes. |
80 | |
77 | |
81 | Special care has been taken to make this module useful from other modules, |
78 | Special care has been taken to make this module useful from other modules, |
82 | while still supporting specialised environments such as L<App::Staticperl> |
79 | while still supporting specialised environments such as L<App::Staticperl> |
83 | or L<PAR::Packer>. |
80 | or L<PAR::Packer>. |
|
|
81 | |
|
|
82 | =head1 WHAT THIS MODULE IS NOT |
|
|
83 | |
|
|
84 | This module only creates processes and lets you pass file handles and |
|
|
85 | strings to it, and run perl code. It does not implement any kind of RPC - |
|
|
86 | there is no back channel from the process back to you, and there is no RPC |
|
|
87 | or message passing going on. |
|
|
88 | |
|
|
89 | If you need some form of RPC, you can either implement it yourself |
|
|
90 | in whatever way you like, use some message-passing module such |
|
|
91 | as L<AnyEvent::MP>, some pipe such as L<AnyEvent::ZeroMQ>, use |
|
|
92 | L<AnyEvent::Handle> on both sides to send e.g. JSON or Storable messages, |
|
|
93 | and so on. |
84 | |
94 | |
85 | =head1 PROBLEM STATEMENT |
95 | =head1 PROBLEM STATEMENT |
86 | |
96 | |
87 | There are two ways to implement parallel processing on UNIX like operating |
97 | There are two ways to implement parallel processing on UNIX like operating |
88 | systems - fork and process, and fork+exec and process. They have different |
98 | systems - fork and process, and fork+exec and process. They have different |
… | |
… | |
244 | use common::sense; |
254 | use common::sense; |
245 | |
255 | |
246 | use Socket (); |
256 | use Socket (); |
247 | |
257 | |
248 | use AnyEvent; |
258 | use AnyEvent; |
249 | use AnyEvent::Fork::Util; |
|
|
250 | use AnyEvent::Util (); |
259 | use AnyEvent::Util (); |
|
|
260 | |
|
|
261 | use IO::FDPass; |
|
|
262 | |
|
|
263 | our $VERSION = 0.2; |
251 | |
264 | |
252 | our $PERL; # the path to the perl interpreter, deduces with various forms of magic |
265 | our $PERL; # the path to the perl interpreter, deduces with various forms of magic |
253 | |
266 | |
254 | =item my $pool = new AnyEvent::Fork key => value... |
267 | =item my $pool = new AnyEvent::Fork key => value... |
255 | |
268 | |
… | |
… | |
272 | |
285 | |
273 | #TODO: maybe append the packet to any existing string command already in the queue |
286 | #TODO: maybe append the packet to any existing string command already in the queue |
274 | |
287 | |
275 | # ideally, we would want to use "a (w/a)*" as format string, but perl versions |
288 | # ideally, we would want to use "a (w/a)*" as format string, but perl versions |
276 | # from at least 5.8.9 to 5.16.3 are all buggy and can't unpack it. |
289 | # from at least 5.8.9 to 5.16.3 are all buggy and can't unpack it. |
277 | push @{ $self->[2] }, pack "N/a", pack "(w/a)*", @_; |
290 | push @{ $self->[2] }, pack "L/a*", pack "(w/a*)*", @_; |
278 | |
291 | |
279 | $self->[3] ||= AE::io $self->[1], 1, sub { |
292 | $self->[3] ||= AE::io $self->[1], 1, sub { |
280 | # send the next "thing" in the queue - either a reference to an fh, |
293 | # send the next "thing" in the queue - either a reference to an fh, |
281 | # or a plain string. |
294 | # or a plain string. |
282 | |
295 | |
283 | if (ref $self->[2][0]) { |
296 | if (ref $self->[2][0]) { |
284 | # send fh |
297 | # send fh |
285 | AnyEvent::Fork::Util::fd_send fileno $self->[1], fileno ${ $self->[2][0] } |
298 | IO::FDPass::send fileno $self->[1], fileno ${ $self->[2][0] } |
286 | and shift @{ $self->[2] }; |
299 | and shift @{ $self->[2] }; |
287 | |
300 | |
288 | } else { |
301 | } else { |
289 | # send string |
302 | # send string |
290 | my $len = syswrite $self->[1], $self->[2][0] |
303 | my $len = syswrite $self->[1], $self->[2][0] |
… | |
… | |
298 | undef $self->[3]; |
311 | undef $self->[3]; |
299 | # invoke run callback |
312 | # invoke run callback |
300 | $self->[0]->($self->[1]) if $self->[0]; |
313 | $self->[0]->($self->[1]) if $self->[0]; |
301 | } |
314 | } |
302 | }; |
315 | }; |
|
|
316 | |
|
|
317 | () # make sure we don't leak the watcher |
303 | } |
318 | } |
304 | |
319 | |
305 | sub _new { |
320 | sub _new { |
306 | my ($self, $fh) = @_; |
321 | my ($self, $fh) = @_; |
307 | |
322 | |
… | |
… | |
327 | if ($pid eq 0) { |
342 | if ($pid eq 0) { |
328 | require AnyEvent::Fork::Serve; |
343 | require AnyEvent::Fork::Serve; |
329 | $AnyEvent::Fork::Serve::OWNER = $parent; |
344 | $AnyEvent::Fork::Serve::OWNER = $parent; |
330 | close $fh; |
345 | close $fh; |
331 | $0 = "$_[1] of $parent"; |
346 | $0 = "$_[1] of $parent"; |
|
|
347 | $SIG{CHLD} = 'IGNORE'; |
332 | AnyEvent::Fork::Serve::serve ($slave); |
348 | AnyEvent::Fork::Serve::serve ($slave); |
333 | AnyEvent::Fork::Util::_exit 0; |
349 | exit 0; |
334 | } elsif (!$pid) { |
350 | } elsif (!$pid) { |
335 | die "AnyEvent::Fork::Early/Template: unable to fork template process: $!"; |
351 | die "AnyEvent::Fork::Early/Template: unable to fork template process: $!"; |
336 | } |
352 | } |
337 | |
353 | |
338 | AnyEvent::Fork->_new ($fh) |
354 | AnyEvent::Fork->_new ($fh) |
… | |
… | |
413 | my $perl = $; |
429 | my $perl = $; |
414 | |
430 | |
415 | # first we try $^X, but the path must be absolute (always on win32), and end in sth. |
431 | # first we try $^X, but the path must be absolute (always on win32), and end in sth. |
416 | # that looks like perl. this obviously only works for posix and win32 |
432 | # that looks like perl. this obviously only works for posix and win32 |
417 | unless ( |
433 | unless ( |
418 | (AnyEvent::Fork::Util::WIN32 || $perl =~ m%^/%) |
434 | ($^O eq "MSWin32" || $perl =~ m%^/%) |
419 | && $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i |
435 | && $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i |
420 | ) { |
436 | ) { |
421 | # if it doesn't look perlish enough, try Config |
437 | # if it doesn't look perlish enough, try Config |
422 | require Config; |
438 | require Config; |
423 | $perl = $Config::Config{perlpath}; |
439 | $perl = $Config::Config{perlpath}; |
… | |
… | |
434 | Proc::FastSpawn::fd_inherit (fileno $fh, 0); |
450 | Proc::FastSpawn::fd_inherit (fileno $fh, 0); |
435 | |
451 | |
436 | # quick. also doesn't work in win32. of course. what did you expect |
452 | # quick. also doesn't work in win32. of course. what did you expect |
437 | #local $ENV{PERL5LIB} = join ":", grep !ref, @INC; |
453 | #local $ENV{PERL5LIB} = join ":", grep !ref, @INC; |
438 | my %env = %ENV; |
454 | my %env = %ENV; |
439 | $env{PERL5LIB} = join +(AnyEvent::Fork::Util::WIN32 ? ";" : ":"), grep !ref, @INC; |
455 | $env{PERL5LIB} = join +($^O eq "MSWin32" ? ";" : ":"), grep !ref, @INC; |
440 | |
456 | |
441 | Proc::FastSpawn::spawn ( |
457 | Proc::FastSpawn::spawn ( |
442 | $perl, |
458 | $perl, |
443 | ["perl", "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$], |
459 | ["perl", "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$], |
444 | [map "$_=$env{$_}", keys %env], |
460 | [map "$_=$env{$_}", keys %env], |
… | |
… | |
601 | $self->_cmd (r => $func); |
617 | $self->_cmd (r => $func); |
602 | } |
618 | } |
603 | |
619 | |
604 | =back |
620 | =back |
605 | |
621 | |
|
|
622 | =head1 PERFORMANCE |
|
|
623 | |
|
|
624 | Now for some unscientific benchmark numbers (all done on an amd64 |
|
|
625 | GNU/Linux box). These are intended to give you an idea of the relative |
|
|
626 | performance you can expect. |
|
|
627 | |
|
|
628 | Ok, so, I ran a simple benchmark that creates a socketpair, forks, calls |
|
|
629 | exit in the child and waits for the socket to close in the parent. I did |
|
|
630 | load AnyEvent, EV and AnyEvent::Fork, for a total process size of 6312kB. |
|
|
631 | |
|
|
632 | 2079 new processes per second, using socketpair + fork manually |
|
|
633 | |
|
|
634 | Then I did the same thing, but instead of calling fork, I called |
|
|
635 | AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the |
|
|
636 | socket form the child to close on exit. This does the same thing as manual |
|
|
637 | socketpair + fork, except that what is forked is the template process |
|
|
638 | (2440kB), and the socket needs to be passed to the server at the other end |
|
|
639 | of the socket first. |
|
|
640 | |
|
|
641 | 2307 new processes per second, using AnyEvent::Fork->new |
|
|
642 | |
|
|
643 | And finally, using C<new_exec> instead C<new>, using vforks+execs to exec |
|
|
644 | a new perl interpreter and compile the small server each time, I get: |
|
|
645 | |
|
|
646 | 479 vfork+execs per second, using AnyEvent::Fork->new_exec |
|
|
647 | |
|
|
648 | So how can C<< AnyEvent->new >> be faster than a standard fork, een though |
|
|
649 | it uses the same operations, but adds a lot of overhead? |
|
|
650 | |
|
|
651 | The difference is simply the process size: forking the 6MB process takes |
|
|
652 | so much longer than forking the 2.5MB template process that the overhead |
|
|
653 | introduced is canceled out. |
|
|
654 | |
|
|
655 | If the benchmark process grows, the normal fork becomes even slower: |
|
|
656 | |
|
|
657 | 1340 new processes, manual fork in a 20MB process |
|
|
658 | 731 new processes, manual fork in a 200MB process |
|
|
659 | 235 new processes, manual fork in a 2000MB process |
|
|
660 | |
|
|
661 | What that means (to me) is that I can use this module without havign a |
|
|
662 | very bad conscience because of the extra overhead requried to strat new |
|
|
663 | processes. |
|
|
664 | |
|
|
665 | =head1 TYPICAL PROBLEMS |
|
|
666 | |
|
|
667 | This section lists typical problems that remain. I hope by recognising |
|
|
668 | them, most can be avoided. |
|
|
669 | |
|
|
670 | =over 4 |
|
|
671 | |
|
|
672 | =item "leaked" file descriptors for exec'ed processes |
|
|
673 | |
|
|
674 | POSIX systems inherit file descriptors by default when exec'ing a new |
|
|
675 | process. While perl itself laudably sets the close-on-exec flags on new |
|
|
676 | file handles, most C libraries don't care, and even if all cared, it's |
|
|
677 | often not possible to set the flag in a race-free manner. |
|
|
678 | |
|
|
679 | That means some file descriptors can leak through. And since it isn't |
|
|
680 | possible to know which file descriptors are "good" and "neccessary" (or |
|
|
681 | even to know which file descreiptors are open), there is no good way to |
|
|
682 | close the ones that might harm. |
|
|
683 | |
|
|
684 | As an example of what "harm" can be done consider a web server that |
|
|
685 | accepts connections and afterwards some module uses AnyEvent::Fork for the |
|
|
686 | first time, causing it to fork and exec a new process, which might inherit |
|
|
687 | the network socket. When the server closes the socket, it is still open |
|
|
688 | in the child (which doesn't even know that) and the client might conclude |
|
|
689 | that the connection is still fine. |
|
|
690 | |
|
|
691 | For the main program, there are multiple remedies available - |
|
|
692 | L<AnyEvent::Fork::Early> is one, creating a process early and not using |
|
|
693 | C<new_exec> is another, as in both cases, the first process can be exec'ed |
|
|
694 | well before many random file descriptors are open. |
|
|
695 | |
|
|
696 | In general, the solution for these kind of problems is to fix the |
|
|
697 | libraries or the code that leaks those file descriptors. |
|
|
698 | |
|
|
699 | Fortunately, most of these lekaed descriptors do no harm, other than |
|
|
700 | sitting on some resources. |
|
|
701 | |
|
|
702 | =item "leaked" file descriptors for fork'ed processes |
|
|
703 | |
|
|
704 | Normally, L<AnyEvent::Fork> does start new processes by exec'ing them, |
|
|
705 | which closes file descriptors not marked for being inherited. |
|
|
706 | |
|
|
707 | However, L<AnyEvent::Fork::Early> and L<AnyEvent::Fork::Template> offer |
|
|
708 | a way to create these processes by forking, and this leaks more file |
|
|
709 | descriptors than exec'ing them, as there is no way to mark descriptors as |
|
|
710 | "close on fork". |
|
|
711 | |
|
|
712 | An example would be modules like L<EV>, L<IO::AIO> or L<Gtk2>. Both create |
|
|
713 | pipes for internal uses, and L<Gtk2> might open a connection to the X |
|
|
714 | server. L<EV> and L<IO::AIO> can deal with fork, but Gtk2 might have |
|
|
715 | trouble with a fork. |
|
|
716 | |
|
|
717 | The solution is to either not load these modules before use'ing |
|
|
718 | L<AnyEvent::Fork::Early> or L<AnyEvent::Fork::Template>, or to delay |
|
|
719 | initialising them, for example, by calling C<init Gtk2> manually. |
|
|
720 | |
|
|
721 | =back |
|
|
722 | |
606 | =head1 PORTABILITY NOTES |
723 | =head1 PORTABILITY NOTES |
607 | |
724 | |
608 | Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop, |
725 | Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop, |
609 | and ::Template is not going to work), and it cost a lot of blood and sweat |
726 | and ::Template is not going to work), and it cost a lot of blood and sweat |
610 | to make it so, mostly due to the bloody broken perl that nobody seems to |
727 | to make it so, mostly due to the bloody broken perl that nobody seems to |
… | |
… | |
614 | |
731 | |
615 | Cygwin perl is not supported at the moment, as it should implement fd |
732 | Cygwin perl is not supported at the moment, as it should implement fd |
616 | passing, but doesn't, and rolling my own is hard, as cygwin doesn't |
733 | passing, but doesn't, and rolling my own is hard, as cygwin doesn't |
617 | support enough functionality to do it. |
734 | support enough functionality to do it. |
618 | |
735 | |
|
|
736 | =head1 SEE ALSO |
|
|
737 | |
|
|
738 | L<AnyEvent::Fork::Early> (to avoid executing a perl interpreter), |
|
|
739 | L<AnyEvent::Fork::Template> (to create a process by forking the main |
|
|
740 | program at a convenient time). |
|
|
741 | |
619 | =head1 AUTHOR |
742 | =head1 AUTHOR |
620 | |
743 | |
621 | Marc Lehmann <schmorp@schmorp.de> |
744 | Marc Lehmann <schmorp@schmorp.de> |
622 | http://home.schmorp.de/ |
745 | http://home.schmorp.de/ |
623 | |
746 | |