| 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 | |
8 | |
| 9 | =head1 DESCRIPTION |
9 | =head1 DESCRIPTION |
| 10 | |
10 | |
| 11 | This module allows you to create single worker processes but also worker |
11 | 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 |
12 | them from your current process (avoiding the problems of forking), but |
| 13 | perl interpreters from a module. |
13 | preserving most of the advantages of fork. |
| 14 | |
14 | |
| 15 | You create a new processes in a pool by specifying a function to call |
15 | It can be used to create new worker processes or new independent |
| 16 | with any combination of string values and file handles. |
16 | subprocesses for short- and long-running jobs, process pools (e.g. for use |
|
|
17 | in pre-forked servers) but also to spawn new external processes (such as |
|
|
18 | CGI scripts from a webserver), which can be faster (and more well behaved) |
|
|
19 | than using fork+exec in big processes. |
| 17 | |
20 | |
| 18 | A pool can have initialisation code which is executed before forking. The |
21 | 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 |
22 | while still supporting specialised environments such as L<App::Staticperl> |
| 20 | cached, to be used as a template. |
23 | or L<PAR::Packer>. |
| 21 | |
|
|
| 22 | Pools without such initialisation code don't cache an extra process. |
|
|
| 23 | |
24 | |
| 24 | =head1 PROBLEM STATEMENT |
25 | =head1 PROBLEM STATEMENT |
| 25 | |
26 | |
| 26 | There are two ways to implement parallel processing on UNIX like operating |
27 | There are two ways to implement parallel processing on UNIX like operating |
| 27 | systems - fork and process, and fork+exec and process. They have different |
28 | systems - fork and process, and fork+exec and process. They have different |
| … | |
… | |
| 145 | |
146 | |
| 146 | =over 4 |
147 | =over 4 |
| 147 | |
148 | |
| 148 | =cut |
149 | =cut |
| 149 | |
150 | |
| 150 | package AnyEvent::ProcessPool; |
151 | package AnyEvent::Fork; |
| 151 | |
152 | |
| 152 | use common::sense; |
153 | use common::sense; |
| 153 | |
154 | |
| 154 | use Socket (); |
155 | use Socket (); |
| 155 | |
156 | |
| 156 | use Proc::FastSpawn; |
|
|
| 157 | use AnyEvent; |
157 | use AnyEvent; |
| 158 | use AnyEvent::ProcessPool::Util; |
158 | use AnyEvent::Fork::Util; |
| 159 | use AnyEvent::Util (); |
159 | use AnyEvent::Util (); |
| 160 | |
160 | |
| 161 | BEGIN { |
161 | our $PERL; # the path to the perl interpreter, deduces with various forms of magic |
| 162 | # require Exporter; |
|
|
| 163 | } |
|
|
| 164 | |
162 | |
| 165 | =item my $pool = new AnyEvent::ProcessPool key => value... |
163 | =item my $pool = new AnyEvent::Fork key => value... |
| 166 | |
164 | |
| 167 | Create a new process pool. The following named parameters are supported: |
165 | Create a new process pool. The following named parameters are supported: |
| 168 | |
166 | |
| 169 | =over 4 |
167 | =over 4 |
| 170 | |
168 | |
| 171 | =back |
169 | =back |
| 172 | |
170 | |
| 173 | =cut |
171 | =cut |
| 174 | |
172 | |
|
|
173 | # the early fork template process |
|
|
174 | our $EARLY; |
|
|
175 | |
| 175 | # the template process |
176 | # the empty template process |
| 176 | our $template; |
177 | our $TEMPLATE; |
| 177 | |
178 | |
| 178 | sub _queue { |
179 | sub _cmd { |
|
|
180 | my $self = shift; |
|
|
181 | |
|
|
182 | # ideally, we would want to use "a (w/a)*" as format string, but perl versions |
|
|
183 | # from at least 5.8.9 to 5.16.3 are all buggy and can't unpack it. |
|
|
184 | push @{ $self->[2] }, pack "N/a", pack "(w/a)*", @_; |
|
|
185 | |
|
|
186 | $self->[3] ||= AE::io $self->[1], 1, sub { |
|
|
187 | if (ref $self->[2][0]) { |
|
|
188 | AnyEvent::Fork::Util::fd_send fileno $self->[1], fileno ${ $self->[2][0] } |
|
|
189 | and shift @{ $self->[2] }; |
|
|
190 | |
|
|
191 | } else { |
|
|
192 | my $len = syswrite $self->[1], $self->[2][0] |
|
|
193 | or do { undef $self->[3]; die "AnyEvent::Fork: command write failure: $!" }; |
|
|
194 | |
|
|
195 | substr $self->[2][0], 0, $len, ""; |
|
|
196 | shift @{ $self->[2] } unless length $self->[2][0]; |
|
|
197 | } |
|
|
198 | |
|
|
199 | unless (@{ $self->[2] }) { |
|
|
200 | undef $self->[3]; |
|
|
201 | $self->[0]->($self->[1]) if $self->[0]; |
|
|
202 | } |
|
|
203 | }; |
|
|
204 | } |
|
|
205 | |
|
|
206 | sub _new { |
| 179 | my ($pid, $fh) = @_; |
207 | my ($self, $fh) = @_; |
| 180 | |
208 | |
| 181 | [ |
209 | AnyEvent::Util::fh_nonblocking $fh, 1; |
| 182 | $pid, |
210 | |
|
|
211 | $self = bless [ |
|
|
212 | undef, # run callback |
| 183 | $fh, |
213 | $fh, |
| 184 | [], |
214 | [], # write queue - strings or fd's |
| 185 | undef |
215 | undef, # AE watcher |
| 186 | ] |
216 | ], $self; |
| 187 | } |
|
|
| 188 | |
217 | |
| 189 | sub queue_cmd { |
218 | # my ($a, $b) = AnyEvent::Util::portable_socketpair; |
| 190 | my $queue = shift; |
|
|
| 191 | |
219 | |
| 192 | push @{ $queue->[2] }, pack "N/a", pack "a (w/a)*", @_; |
220 | # queue_cmd $template, "Iabc"; |
|
|
221 | # push @{ $template->[2] }, \$b; |
| 193 | |
222 | |
| 194 | $queue->[3] ||= AE::io $queue->[1], 1, sub { |
223 | # use Coro::AnyEvent; Coro::AnyEvent::sleep 1; |
| 195 | if (ref $queue->[2][0]) { |
224 | # undef $b; |
| 196 | AnyEvent::ProcessPool::Util::fd_send fileno $queue->[1], fileno ${ $queue->[2][0] } |
225 | # die "x" . <$a>; |
| 197 | and shift @{ $queue->[2] }; |
|
|
| 198 | } else { |
|
|
| 199 | my $len = syswrite $queue->[1], $queue->[2][0] |
|
|
| 200 | or do { undef $queue->[3]; die "AnyEvent::ProcessPool::queue write failure: $!" }; |
|
|
| 201 | substr $queue->[2][0], 0, $len, ""; |
|
|
| 202 | shift @{ $queue->[2] } unless length $queue->[2][0]; |
|
|
| 203 | } |
|
|
| 204 | |
226 | |
| 205 | undef $queue->[3] unless @{ $queue->[2] }; |
227 | $self |
| 206 | }; |
|
|
| 207 | } |
228 | } |
| 208 | |
229 | |
| 209 | sub run_template { |
230 | # fork template from current process, used by AnyEvent::Fork::Early/Template |
| 210 | return if $template; |
231 | sub _new_fork { |
| 211 | |
|
|
| 212 | my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
232 | my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
| 213 | AnyEvent::Util::fh_nonblocking $fh, 1; |
233 | my $parent = $$; |
| 214 | fd_inherit fileno $slave; |
|
|
| 215 | |
234 | |
|
|
235 | my $pid = fork; |
|
|
236 | |
|
|
237 | if ($pid eq 0) { |
|
|
238 | require AnyEvent::Fork::Serve; |
|
|
239 | $AnyEvent::Fork::Serve::OWNER = $parent; |
|
|
240 | close $fh; |
|
|
241 | $0 = "$_[1] of $parent"; |
|
|
242 | AnyEvent::Fork::Serve::serve ($slave); |
|
|
243 | AnyEvent::Fork::Util::_exit 0; |
|
|
244 | } elsif (!$pid) { |
|
|
245 | die "AnyEvent::Fork::Early/Template: unable to fork template process: $!"; |
|
|
246 | } |
|
|
247 | |
|
|
248 | AnyEvent::Fork->_new ($fh) |
|
|
249 | } |
|
|
250 | |
|
|
251 | =item my $proc = new AnyEvent::Fork |
|
|
252 | |
|
|
253 | Create a new "empty" perl interpreter process and returns its process |
|
|
254 | object for further manipulation. |
|
|
255 | |
|
|
256 | The new process is forked from a template process that is kept around |
|
|
257 | for this purpose. When it doesn't exist yet, it is created by a call to |
|
|
258 | C<new_exec> and kept around for future calls. |
|
|
259 | |
|
|
260 | =cut |
|
|
261 | |
|
|
262 | sub new { |
|
|
263 | my $class = shift; |
|
|
264 | |
|
|
265 | $TEMPLATE ||= $class->new_exec; |
|
|
266 | $TEMPLATE->fork |
|
|
267 | } |
|
|
268 | |
|
|
269 | =item $new_proc = $proc->fork |
|
|
270 | |
|
|
271 | Forks C<$proc>, creating a new process, and returns the process object |
|
|
272 | of the new process. |
|
|
273 | |
|
|
274 | If any of the C<send_> functions have been called before fork, then they |
|
|
275 | will be cloned in the child. For example, in a pre-forked server, you |
|
|
276 | might C<send_fh> the listening socket into the template process, and then |
|
|
277 | keep calling C<fork> and C<run>. |
|
|
278 | |
|
|
279 | =cut |
|
|
280 | |
|
|
281 | sub fork { |
|
|
282 | my ($self) = @_; |
|
|
283 | |
|
|
284 | my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
|
|
285 | |
|
|
286 | $self->send_fh ($slave); |
|
|
287 | $self->_cmd ("f"); |
|
|
288 | |
|
|
289 | AnyEvent::Fork->_new ($fh) |
|
|
290 | } |
|
|
291 | |
|
|
292 | =item my $proc = new_exec AnyEvent::Fork |
|
|
293 | |
|
|
294 | Create a new "empty" perl interpreter process and returns its process |
|
|
295 | object for further manipulation. |
|
|
296 | |
|
|
297 | Unlike the C<new> method, this method I<always> spawns a new perl process |
|
|
298 | (except in some cases, see L<AnyEvent::Fork::Early> for details). This |
|
|
299 | reduces the amount of memory sharing that is possible, and is also slower. |
|
|
300 | |
|
|
301 | You should use C<new> whenever possible, except when having a template |
|
|
302 | process around is unacceptable. |
|
|
303 | |
|
|
304 | The path to the perl interpreter is divined usign various methods - first |
|
|
305 | C<$^X> is investigated to see if the path ends with something that sounds |
|
|
306 | as if it were the perl interpreter. Failing this, the module falls back to |
|
|
307 | using C<$Config::Config{perlpath}>. |
|
|
308 | |
|
|
309 | =cut |
|
|
310 | |
|
|
311 | sub new_exec { |
|
|
312 | my ($self) = @_; |
|
|
313 | |
|
|
314 | return $EARLY->fork |
|
|
315 | if $EARLY; |
|
|
316 | |
|
|
317 | # first find path of perl |
|
|
318 | my $perl = $; |
|
|
319 | |
|
|
320 | # first we try $^X, but the path must be absolute (always on win32), and end in sth. |
|
|
321 | # that looks like perl. this obviously only works for posix and win32 |
|
|
322 | unless ( |
|
|
323 | (AnyEvent::Fork::Util::WIN32 || $perl =~ m%^/%) |
|
|
324 | && $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i |
|
|
325 | ) { |
|
|
326 | # if it doesn't look perlish enough, try Config |
|
|
327 | require Config; |
|
|
328 | $perl = $Config::Config{perlpath}; |
|
|
329 | $perl =~ s/(?:\Q$Config::Config{_exe}\E)?$/$Config::Config{_exe}/; |
|
|
330 | } |
|
|
331 | |
|
|
332 | require Proc::FastSpawn; |
|
|
333 | |
|
|
334 | my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
|
|
335 | Proc::FastSpawn::fd_inherit (fileno $slave); |
|
|
336 | |
|
|
337 | # quick. also doesn't work in win32. of course. what did you expect |
|
|
338 | #local $ENV{PERL5LIB} = join ":", grep !ref, @INC; |
| 216 | my %env = %ENV; |
339 | my %env = %ENV; |
| 217 | $env{PERL5LIB} = join ":", grep !ref, @INC; |
340 | $env{PERL5LIB} = join ":", grep !ref, @INC; |
| 218 | |
341 | |
| 219 | my $pid = spawn |
342 | Proc::FastSpawn::spawn ( |
| 220 | $^X, |
343 | $perl, |
| 221 | ["perl", "-MAnyEvent::ProcessPool::Serve", "-e", "AnyEvent::ProcessPool::Serve::me", fileno $slave], |
344 | ["perl", "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$], |
| 222 | [map "$_=$env{$_}", keys %env], |
345 | [map "$_=$env{$_}", keys %env], |
| 223 | or die "unable to spawn AnyEvent::ProcessPool server: $!"; |
346 | ) or die "unable to spawn AnyEvent::Fork server: $!"; |
| 224 | |
347 | |
| 225 | close $slave; |
348 | $self->_new ($fh) |
|
|
349 | } |
| 226 | |
350 | |
| 227 | $template = _queue $pid, $fh; |
351 | =item $proc = $proc->require ($module, ...) |
| 228 | |
352 | |
| 229 | my ($a, $b) = AnyEvent::Util::portable_socketpair; |
353 | Tries to load the given modules into the process |
| 230 | |
354 | |
| 231 | queue_cmd $template, "Iabc"; |
355 | Returns the process object for easy chaining of method calls. |
|
|
356 | |
|
|
357 | =item $proc = $proc->send_fh ($handle, ...) |
|
|
358 | |
|
|
359 | Send one or more file handles (I<not> file descriptors) to the process, |
|
|
360 | to prepare a call to C<run>. |
|
|
361 | |
|
|
362 | The process object keeps a reference to the handles until this is done, |
|
|
363 | so you must not explicitly close the handles. This is most easily |
|
|
364 | accomplished by simply not storing the file handles anywhere after passing |
|
|
365 | them to this method. |
|
|
366 | |
|
|
367 | Returns the process object for easy chaining of method calls. |
|
|
368 | |
|
|
369 | =cut |
|
|
370 | |
|
|
371 | sub send_fh { |
|
|
372 | my ($self, @fh) = @_; |
|
|
373 | |
|
|
374 | for my $fh (@fh) { |
|
|
375 | $self->_cmd ("h"); |
| 232 | push @{ $template->[2] }, \$b; |
376 | push @{ $self->[2] }, \$fh; |
| 233 | |
377 | } |
| 234 | use Coro::AnyEvent; Coro::AnyEvent::sleep 1; |
|
|
| 235 | undef $b; |
|
|
| 236 | die "x" . <$a>; |
|
|
| 237 | } |
|
|
| 238 | |
|
|
| 239 | sub new { |
|
|
| 240 | my $class = shift; |
|
|
| 241 | |
|
|
| 242 | my $self = bless { |
|
|
| 243 | @_ |
|
|
| 244 | }, $class; |
|
|
| 245 | |
|
|
| 246 | run_template; |
|
|
| 247 | |
378 | |
| 248 | $self |
379 | $self |
|
|
380 | } |
|
|
381 | |
|
|
382 | =item $proc = $proc->send_arg ($string, ...) |
|
|
383 | |
|
|
384 | Send one or more argument strings to the process, to prepare a call to |
|
|
385 | C<run>. The strings can be any octet string. |
|
|
386 | |
|
|
387 | Returns the process object for easy chaining of emthod calls. |
|
|
388 | |
|
|
389 | =cut |
|
|
390 | |
|
|
391 | sub send_arg { |
|
|
392 | my ($self, @arg) = @_; |
|
|
393 | |
|
|
394 | $self->_cmd (a => @arg); |
|
|
395 | |
|
|
396 | $self |
|
|
397 | } |
|
|
398 | |
|
|
399 | =item $proc->run ($func, $cb->($fh)) |
|
|
400 | |
|
|
401 | Enter the function specified by the fully qualified name in C<$func> in |
|
|
402 | the process. The function is called with the communication socket as first |
|
|
403 | argument, followed by all file handles and string arguments sent earlier |
|
|
404 | via C<send_fh> and C<send_arg> methods, in the order they were called. |
|
|
405 | |
|
|
406 | If the called function returns, the process exits. |
|
|
407 | |
|
|
408 | Preparing the process can take time - when the process is ready, the |
|
|
409 | callback is invoked with the local communications socket as argument. |
|
|
410 | |
|
|
411 | The process object becomes unusable on return from this function. |
|
|
412 | |
|
|
413 | If the communication socket isn't used, it should be closed on both sides, |
|
|
414 | to save on kernel memory. |
|
|
415 | |
|
|
416 | The socket is non-blocking in the parent, and blocking in the newly |
|
|
417 | created process. The close-on-exec flag is set on both. Even if not used |
|
|
418 | otherwise, the socket can be a good indicator for the existance of the |
|
|
419 | process - if the othe rprocess exits, you get a readable event on it, |
|
|
420 | because exiting the process closes the socket (if it didn't create any |
|
|
421 | children using fork). |
|
|
422 | |
|
|
423 | =cut |
|
|
424 | |
|
|
425 | sub run { |
|
|
426 | my ($self, $func, $cb) = @_; |
|
|
427 | |
|
|
428 | $self->[0] = $cb; |
|
|
429 | $self->_cmd ("r", $func); |
| 249 | } |
430 | } |
| 250 | |
431 | |
| 251 | =back |
432 | =back |
| 252 | |
433 | |
| 253 | =head1 AUTHOR |
434 | =head1 AUTHOR |
