ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent-Fork/README
(Generate patch)

Comparing AnyEvent-Fork/README (file contents):
Revision 1.3 by root, Fri Apr 5 19:10:10 2013 UTC vs.
Revision 1.6 by root, Thu Apr 18 20:17:35 2013 UTC

… …
2 AnyEvent::Fork - everything you wanted to use fork() for, but couldn't 2 AnyEvent::Fork - everything you wanted to use fork() for, but couldn't
3 3
4SYNOPSIS 4SYNOPSIS
5 use AnyEvent::Fork; 5 use AnyEvent::Fork;
6 6
7 ################################################################## 7 AnyEvent::Fork
8 ->new
9 ->require ("MyModule")
10 ->run ("MyModule::server", my $cv = AE::cv);
11
12 my $fh = $cv->recv;
13
14DESCRIPTION
15 This module allows you to create new processes, without actually forking
16 them from your current process (avoiding the problems of forking), but
17 preserving most of the advantages of fork.
18
19 It can be used to create new worker processes or new independent
20 subprocesses for short- and long-running jobs, process pools (e.g. for
21 use in pre-forked servers) but also to spawn new external processes
22 (such as CGI scripts from a web server), which can be faster (and more
23 well behaved) than using fork+exec in big processes.
24
25 Special care has been taken to make this module useful from other
26 modules, while still supporting specialised environments such as
27 App::Staticperl or PAR::Packer.
28
29 WHAT THIS MODULE IS NOT
30 This module only creates processes and lets you pass file handles and
31 strings to it, and run perl code. It does not implement any kind of RPC
32 - there is no back channel from the process back to you, and there is no
33 RPC or message passing going on.
34
35 If you need some form of RPC, you could use the AnyEvent::Fork::RPC
36 companion module, which adds simple RPC/job queueing to a process
37 created by this module.
38
39 Or you can implement it yourself in whatever way you like, use some
40 message-passing module such as AnyEvent::MP, some pipe such as
41 AnyEvent::ZeroMQ, use AnyEvent::Handle on both sides to send e.g. JSON
42 or Storable messages, and so on.
43
44 COMPARISON TO OTHER MODULES
45 There is an abundance of modules on CPAN that do "something fork", such
46 as Parallel::ForkManager, AnyEvent::ForkManager, AnyEvent::Worker or
47 AnyEvent::Subprocess. There are modules that implement their own process
48 management, such as AnyEvent::DBI.
49
50 The problems that all these modules try to solve are real, however, none
51 of them (from what I have seen) tackle the very real problems of
52 unwanted memory sharing, efficiency, not being able to use event
53 processing or similar modules in the processes they create.
54
55 This module doesn't try to replace any of them - instead it tries to
56 solve the problem of creating processes with a minimum of fuss and
57 overhead (and also luxury). Ideally, most of these would use
58 AnyEvent::Fork internally, except they were written before AnyEvent:Fork
59 was available, so obviously had to roll their own.
60
61 PROBLEM STATEMENT
62 There are two traditional ways to implement parallel processing on UNIX
63 like operating systems - fork and process, and fork+exec and process.
64 They have different advantages and disadvantages that I describe below,
65 together with how this module tries to mitigate the disadvantages.
66
67 Forking from a big process can be very slow.
68 A 5GB process needs 0.05s to fork on my 3.6GHz amd64 GNU/Linux box.
69 This overhead is often shared with exec (because you have to fork
70 first), but in some circumstances (e.g. when vfork is used),
71 fork+exec can be much faster.
72
73 This module can help here by telling a small(er) helper process to
74 fork, which is faster then forking the main process, and also uses
75 vfork where possible. This gives the speed of vfork, with the
76 flexibility of fork.
77
78 Forking usually creates a copy-on-write copy of the parent process.
79 For example, modules or data files that are loaded will not use
80 additional memory after a fork. When exec'ing a new process, modules
81 and data files might need to be loaded again, at extra CPU and
82 memory cost. But when forking, literally all data structures are
83 copied - if the program frees them and replaces them by new data,
84 the child processes will retain the old version even if it isn't
85 used, which can suddenly and unexpectedly increase memory usage when
86 freeing memory.
87
88 The trade-off is between more sharing with fork (which can be good
89 or bad), and no sharing with exec.
90
91 This module allows the main program to do a controlled fork, and
92 allows modules to exec processes safely at any time. When creating a
93 custom process pool you can take advantage of data sharing via fork
94 without risking to share large dynamic data structures that will
95 blow up child memory usage.
96
97 In other words, this module puts you into control over what is being
98 shared and what isn't, at all times.
99
100 Exec'ing a new perl process might be difficult.
101 For example, it is not easy to find the correct path to the perl
102 interpreter - $^X might not be a perl interpreter at all.
103
104 This module tries hard to identify the correct path to the perl
105 interpreter. With a cooperative main program, exec'ing the
106 interpreter might not even be necessary, but even without help from
107 the main program, it will still work when used from a module.
108
109 Exec'ing a new perl process might be slow, as all necessary modules have
110 to be loaded from disk again, with no guarantees of success.
111 Long running processes might run into problems when perl is upgraded
112 and modules are no longer loadable because they refer to a different
113 perl version, or parts of a distribution are newer than the ones
114 already loaded.
115
116 This module supports creating pre-initialised perl processes to be
117 used as a template for new processes.
118
119 Forking might be impossible when a program is running.
120 For example, POSIX makes it almost impossible to fork from a
121 multi-threaded program while doing anything useful in the child - in
122 fact, if your perl program uses POSIX threads (even indirectly via
123 e.g. IO::AIO or threads), you cannot call fork on the perl level
124 anymore without risking corruption issues on a number of operating
125 systems.
126
127 This module can safely fork helper processes at any time, by calling
128 fork+exec in C, in a POSIX-compatible way (via Proc::FastSpawn).
129
130 Parallel processing with fork might be inconvenient or difficult to
131 implement. Modules might not work in both parent and child.
132 For example, when a program uses an event loop and creates watchers
133 it becomes very hard to use the event loop from a child program, as
134 the watchers already exist but are only meaningful in the parent.
135 Worse, a module might want to use such a module, not knowing whether
136 another module or the main program also does, leading to problems.
137
138 Apart from event loops, graphical toolkits also commonly fall into
139 the "unsafe module" category, or just about anything that
140 communicates with the external world, such as network libraries and
141 file I/O modules, which usually don't like being copied and then
142 allowed to continue in two processes.
143
144 With this module only the main program is allowed to create new
145 processes by forking (because only the main program can know when it
146 is still safe to do so) - all other processes are created via
147 fork+exec, which makes it possible to use modules such as event
148 loops or window interfaces safely.
149
150EXAMPLES
8 # create a single new process, tell it to run your worker function 151 Create a single new process, tell it to run your worker function.
9
10 AnyEvent::Fork 152 AnyEvent::Fork
11 ->new 153 ->new
12 ->require ("MyModule") 154 ->require ("MyModule")
13 ->run ("MyModule::worker, sub { 155 ->run ("MyModule::worker, sub {
14 my ($master_filehandle) = @_; 156 my ($master_filehandle) = @_;
15 157
16 # now $master_filehandle is connected to the 158 # now $master_filehandle is connected to the
17 # $slave_filehandle in the new process. 159 # $slave_filehandle in the new process.
18 }); 160 });
19 161
20 # MyModule::worker might look like this 162 "MyModule" might look like this:
163
164 package MyModule;
165
21 sub MyModule::worker { 166 sub worker {
22 my ($slave_filehandle) = @_; 167 my ($slave_filehandle) = @_;
23 168
24 # now $slave_filehandle is connected to the $master_filehandle 169 # now $slave_filehandle is connected to the $master_filehandle
25 # in the original prorcess. have fun! 170 # in the original prorcess. have fun!
26 } 171 }
27 172
28 ##################################################################
29 # create a pool of server processes all accepting on the same socket 173 Create a pool of server processes all accepting on the same socket.
30
31 # create listener socket 174 # create listener socket
32 my $listener = ...; 175 my $listener = ...;
33 176
34 # create a pool template, initialise it and give it the socket 177 # create a pool template, initialise it and give it the socket
35 my $pool = AnyEvent::Fork 178 my $pool = AnyEvent::Fork
… …
46 } 189 }
47 190
48 # now do other things - maybe use the filehandle provided by run 191 # now do other things - maybe use the filehandle provided by run
49 # to wait for the processes to die. or whatever. 192 # to wait for the processes to die. or whatever.
50 193
51 # My::Server::run might look like this 194 "My::Server" might look like this:
52 sub My::Server::run { 195
196 package My::Server;
197
198 sub run {
53 my ($slave, $listener, $id) = @_; 199 my ($slave, $listener, $id) = @_;
54 200
55 close $slave; # we do not use the socket, so close it to save resources 201 close $slave; # we do not use the socket, so close it to save resources
56 202
57 # we could go ballistic and use e.g. AnyEvent here, or IO::AIO, 203 # we could go ballistic and use e.g. AnyEvent here, or IO::AIO,
… …
59 while (my $socket = $listener->accept) { 205 while (my $socket = $listener->accept) {
60 # do sth. with new socket 206 # do sth. with new socket
61 } 207 }
62 } 208 }
63 209
64DESCRIPTION 210 use AnyEvent::Fork as a faster fork+exec
65 This module allows you to create new processes, without actually forking 211 This runs "/bin/echo hi", with standard output redirected to /tmp/log
66 them from your current process (avoiding the problems of forking), but 212 and standard error redirected to the communications socket. It is
67 preserving most of the advantages of fork. 213 usually faster than fork+exec, but still lets you prepare the
214 environment.
68 215
69 It can be used to create new worker processes or new independent 216 open my $output, ">/tmp/log" or die "$!";
70 subprocesses for short- and long-running jobs, process pools (e.g. for
71 use in pre-forked servers) but also to spawn new external processes
72 (such as CGI scripts from a webserver), which can be faster (and more
73 well behaved) than using fork+exec in big processes.
74 217
75 Special care has been taken to make this module useful from other 218 AnyEvent::Fork
76 modules, while still supporting specialised environments such as 219 ->new
77 App::Staticperl or PAR::Packer. 220 ->eval ('
221 # compile a helper function for later use
222 sub run {
223 my ($fh, $output, @cmd) = @_;
78 224
79PROBLEM STATEMENT 225 # perl will clear close-on-exec on STDOUT/STDERR
80 There are two ways to implement parallel processing on UNIX like 226 open STDOUT, ">&", $output or die;
81 operating systems - fork and process, and fork+exec and process. They 227 open STDERR, ">&", $fh or die;
82 have different advantages and disadvantages that I describe below,
83 together with how this module tries to mitigate the disadvantages.
84 228
85 Forking from a big process can be very slow (a 5GB process needs 0.05s 229 exec @cmd;
86 to fork on my 3.6GHz amd64 GNU/Linux box for example). This overhead is 230 }
87 often shared with exec (because you have to fork first), but in some 231 ')
88 circumstances (e.g. when vfork is used), fork+exec can be much faster. 232 ->send_fh ($output)
89 This module can help here by telling a small(er) helper process to 233 ->send_arg ("/bin/echo", "hi")
90 fork, or fork+exec instead. 234 ->run ("run", my $cv = AE::cv);
91 235
92 Forking usually creates a copy-on-write copy of the parent process. 236 my $stderr = $cv->recv;
93 Memory (for example, modules or data files that have been will not take
94 additional memory). When exec'ing a new process, modules and data files
95 might need to be loaded again, at extra cpu and memory cost. Likewise
96 when forking, all data structures are copied as well - if the program
97 frees them and replaces them by new data, the child processes will
98 retain the memory even if it isn't used.
99 This module allows the main program to do a controlled fork, and
100 allows modules to exec processes safely at any time. When creating a
101 custom process pool you can take advantage of data sharing via fork
102 without risking to share large dynamic data structures that will
103 blow up child memory usage.
104
105 Exec'ing a new perl process might be difficult and slow. For example, it
106 is not easy to find the correct path to the perl interpreter, and all
107 modules have to be loaded from disk again. Long running processes might
108 run into problems when perl is upgraded for example.
109 This module supports creating pre-initialised perl processes to be
110 used as template, and also tries hard to identify the correct path
111 to the perl interpreter. With a cooperative main program, exec'ing
112 the interpreter might not even be necessary.
113
114 Forking might be impossible when a program is running. For example,
115 POSIX makes it almost impossible to fork from a multithreaded program
116 and do anything useful in the child - strictly speaking, if your perl
117 program uses posix threads (even indirectly via e.g. IO::AIO or
118 threads), you cannot call fork on the perl level anymore, at all.
119 This module can safely fork helper processes at any time, by caling
120 fork+exec in C, in a POSIX-compatible way.
121
122 Parallel processing with fork might be inconvenient or difficult to
123 implement. For example, when a program uses an event loop and creates
124 watchers it becomes very hard to use the event loop from a child
125 program, as the watchers already exist but are only meaningful in the
126 parent. Worse, a module might want to use such a system, not knowing
127 whether another module or the main program also does, leading to
128 problems.
129 This module only lets the main program create pools by forking
130 (because only the main program can know when it is still safe to do
131 so) - all other pools are created by fork+exec, after which such
132 modules can again be loaded.
133 237
134CONCEPTS 238CONCEPTS
135 This module can create new processes either by executing a new perl 239 This module can create new processes either by executing a new perl
136 process, or by forking from an existing "template" process. 240 process, or by forking from an existing "template" process.
241
242 All these processes are called "child processes" (whether they are
243 direct children or not), while the process that manages them is called
244 the "parent process".
137 245
138 Each such process comes with its own file handle that can be used to 246 Each such process comes with its own file handle that can be used to
139 communicate with it (it's actually a socket - one end in the new 247 communicate with it (it's actually a socket - one end in the new
140 process, one end in the main process), and among the things you can do 248 process, one end in the main process), and among the things you can do
141 in it are load modules, fork new processes, send file handles to it, and 249 in it are load modules, fork new processes, send file handles to it, and
… …
151 memory used for the perl interpreter with the new process, but 259 memory used for the perl interpreter with the new process, but
152 loading modules takes time, and the memory is not shared with 260 loading modules takes time, and the memory is not shared with
153 anything else. 261 anything else.
154 262
155 This is ideal for when you only need one extra process of a kind, 263 This is ideal for when you only need one extra process of a kind,
156 with the option of starting and stipping it on demand. 264 with the option of starting and stopping it on demand.
157 265
158 Example: 266 Example:
159 267
160 AnyEvent::Fork 268 AnyEvent::Fork
161 ->new 269 ->new
… …
175 the modules you loaded) is shared between the processes, and each 283 the modules you loaded) is shared between the processes, and each
176 new process consumes relatively little memory of its own. 284 new process consumes relatively little memory of its own.
177 285
178 The disadvantage of this approach is that you need to create a 286 The disadvantage of this approach is that you need to create a
179 template process for the sole purpose of forking new processes from 287 template process for the sole purpose of forking new processes from
180 it, but if you only need a fixed number of proceses you can create 288 it, but if you only need a fixed number of processes you can create
181 them, and then destroy the template process. 289 them, and then destroy the template process.
182 290
183 Example: 291 Example:
184 292
185 my $template = AnyEvent::Fork->new->require ("Some::Module"); 293 my $template = AnyEvent::Fork->new->require ("Some::Module");
… …
209 ->require ("Some::Module") 317 ->require ("Some::Module")
210 ->run ("Some::Module::run", sub { 318 ->run ("Some::Module::run", sub {
211 my ($fork_fh) = @_; 319 my ($fork_fh) = @_;
212 }); 320 });
213 321
214FUNCTIONS 322THE "AnyEvent::Fork" CLASS
215 my $pool = new AnyEvent::Fork key => value... 323 This module exports nothing, and only implements a single class -
216 Create a new process pool. The following named parameters are 324 "AnyEvent::Fork".
217 supported: 325
326 There are two class constructors that both create new processes - "new"
327 and "new_exec". The "fork" method creates a new process by forking an
328 existing one and could be considered a third constructor.
329
330 Most of the remaining methods deal with preparing the new process, by
331 loading code, evaluating code and sending data to the new process. They
332 usually return the process object, so you can chain method calls.
333
334 If a process object is destroyed before calling its "run" method, then
335 the process simply exits. After "run" is called, all responsibility is
336 passed to the specified function.
337
338 As long as there is any outstanding work to be done, process objects
339 resist being destroyed, so there is no reason to store them unless you
340 need them later - configure and forget works just fine.
218 341
219 my $proc = new AnyEvent::Fork 342 my $proc = new AnyEvent::Fork
220 Create a new "empty" perl interpreter process and returns its 343 Create a new "empty" perl interpreter process and returns its
221 process object for further manipulation. 344 process object for further manipulation.
222 345
223 The new process is forked from a template process that is kept 346 The new process is forked from a template process that is kept
224 around for this purpose. When it doesn't exist yet, it is created by 347 around for this purpose. When it doesn't exist yet, it is created by
225 a call to "new_exec" and kept around for future calls. 348 a call to "new_exec" first and then stays around for future calls.
226
227 When the process object is destroyed, it will release the file
228 handle that connects it with the new process. When the new process
229 has not yet called "run", then the process will exit. Otherwise,
230 what happens depends entirely on the code that is executed.
231 349
232 $new_proc = $proc->fork 350 $new_proc = $proc->fork
233 Forks $proc, creating a new process, and returns the process object 351 Forks $proc, creating a new process, and returns the process object
234 of the new process. 352 of the new process.
235 353
… …
248 possible, and is also slower. 366 possible, and is also slower.
249 367
250 You should use "new" whenever possible, except when having a 368 You should use "new" whenever possible, except when having a
251 template process around is unacceptable. 369 template process around is unacceptable.
252 370
253 The path to the perl interpreter is divined usign various methods - 371 The path to the perl interpreter is divined using various methods -
254 first $^X is investigated to see if the path ends with something 372 first $^X is investigated to see if the path ends with something
255 that sounds as if it were the perl interpreter. Failing this, the 373 that sounds as if it were the perl interpreter. Failing this, the
256 module falls back to using $Config::Config{perlpath}. 374 module falls back to using $Config::Config{perlpath}.
257 375
376 $pid = $proc->pid
377 Returns the process id of the process *iff it is a direct child of
378 the process running AnyEvent::Fork*, and "undef" otherwise.
379
380 Normally, only processes created via "AnyEvent::Fork->new_exec" and
381 AnyEvent::Fork::Template are direct children, and you are
382 responsible to clean up their zombies when they die.
383
384 All other processes are not direct children, and will be cleaned up
385 by AnyEvent::Fork itself.
386
258 $proc = $proc->eval ($perlcode, @args) 387 $proc = $proc->eval ($perlcode, @args)
259 Evaluates the given $perlcode as ... perl code, while setting @_ to 388 Evaluates the given $perlcode as ... Perl code, while setting @_ to
260 the strings specified by @args. 389 the strings specified by @args, in the "main" package.
261 390
262 This call is meant to do any custom initialisation that might be 391 This call is meant to do any custom initialisation that might be
263 required (for example, the "require" method uses it). It's not 392 required (for example, the "require" method uses it). It's not
264 supposed to be used to completely take over the process, use "run" 393 supposed to be used to completely take over the process, use "run"
265 for that. 394 for that.
… …
267 The code will usually be executed after this call returns, and there 396 The code will usually be executed after this call returns, and there
268 is no way to pass anything back to the calling process. Any 397 is no way to pass anything back to the calling process. Any
269 evaluation errors will be reported to stderr and cause the process 398 evaluation errors will be reported to stderr and cause the process
270 to exit. 399 to exit.
271 400
401 If you want to execute some code (that isn't in a module) to take
402 over the process, you should compile a function via "eval" first,
403 and then call it via "run". This also gives you access to any
404 arguments passed via the "send_xxx" methods, such as file handles.
405 See the "use AnyEvent::Fork as a faster fork+exec" example to see it
406 in action.
407
272 Returns the process object for easy chaining of method calls. 408 Returns the process object for easy chaining of method calls.
273 409
274 $proc = $proc->require ($module, ...) 410 $proc = $proc->require ($module, ...)
275 Tries to load the given module(s) into the process 411 Tries to load the given module(s) into the process
276 412
… …
278 414
279 $proc = $proc->send_fh ($handle, ...) 415 $proc = $proc->send_fh ($handle, ...)
280 Send one or more file handles (*not* file descriptors) to the 416 Send one or more file handles (*not* file descriptors) to the
281 process, to prepare a call to "run". 417 process, to prepare a call to "run".
282 418
283 The process object keeps a reference to the handles until this is 419 The process object keeps a reference to the handles until they have
284 done, so you must not explicitly close the handles. This is most 420 been passed over to the process, so you must not explicitly close
285 easily accomplished by simply not storing the file handles anywhere 421 the handles. This is most easily accomplished by simply not storing
286 after passing them to this method. 422 the file handles anywhere after passing them to this method - when
423 AnyEvent::Fork is finished using them, perl will automatically close
424 them.
287 425
288 Returns the process object for easy chaining of method calls. 426 Returns the process object for easy chaining of method calls.
289 427
290 Example: pass an fh to a process, and release it without closing. it 428 Example: pass a file handle to a process, and release it without
291 will be closed automatically when it is no longer used. 429 closing. It will be closed automatically when it is no longer used.
292 430
293 $proc->send_fh ($my_fh); 431 $proc->send_fh ($my_fh);
294 undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT 432 undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT
295 433
296 $proc = $proc->send_arg ($string, ...) 434 $proc = $proc->send_arg ($string, ...)
297 Send one or more argument strings to the process, to prepare a call 435 Send one or more argument strings to the process, to prepare a call
298 to "run". The strings can be any octet string. 436 to "run". The strings can be any octet strings.
299 437
438 The protocol is optimised to pass a moderate number of relatively
439 short strings - while you can pass up to 4GB of data in one go, this
440 is more meant to pass some ID information or other startup info, not
441 big chunks of data.
442
300 Returns the process object for easy chaining of emthod calls. 443 Returns the process object for easy chaining of method calls.
301 444
302 $proc->run ($func, $cb->($fh)) 445 $proc->run ($func, $cb->($fh))
303 Enter the function specified by the fully qualified name in $func in 446 Enter the function specified by the function name in $func in the
304 the process. The function is called with the communication socket as 447 process. The function is called with the communication socket as
305 first argument, followed by all file handles and string arguments 448 first argument, followed by all file handles and string arguments
306 sent earlier via "send_fh" and "send_arg" methods, in the order they 449 sent earlier via "send_fh" and "send_arg" methods, in the order they
307 were called. 450 were called.
308 451
309 If the called function returns, the process exits.
310
311 Preparing the process can take time - when the process is ready, the
312 callback is invoked with the local communications socket as
313 argument.
314
315 The process object becomes unusable on return from this function. 452 The process object becomes unusable on return from this function -
453 any further method calls result in undefined behaviour.
454
455 The function name should be fully qualified, but if it isn't, it
456 will be looked up in the "main" package.
457
458 If the called function returns, doesn't exist, or any error occurs,
459 the process exits.
460
461 Preparing the process is done in the background - when all commands
462 have been sent, the callback is invoked with the local
463 communications socket as argument. At this point you can start using
464 the socket in any way you like.
316 465
317 If the communication socket isn't used, it should be closed on both 466 If the communication socket isn't used, it should be closed on both
318 sides, to save on kernel memory. 467 sides, to save on kernel memory.
319 468
320 The socket is non-blocking in the parent, and blocking in the newly 469 The socket is non-blocking in the parent, and blocking in the newly
321 created process. The close-on-exec flag is set on both. Even if not 470 created process. The close-on-exec flag is set in both.
471
322 used otherwise, the socket can be a good indicator for the existance 472 Even if not used otherwise, the socket can be a good indicator for
323 of the process - if the other process exits, you get a readable 473 the existence of the process - if the other process exits, you get a
324 event on it, because exiting the process closes the socket (if it 474 readable event on it, because exiting the process closes the socket
325 didn't create any children using fork). 475 (if it didn't create any children using fork).
326 476
327 Example: create a template for a process pool, pass a few strings, 477 Example: create a template for a process pool, pass a few strings,
328 some file handles, then fork, pass one more string, and run some 478 some file handles, then fork, pass one more string, and run some
329 code. 479 code.
330 480
… …
339 ->send_arg ("str3") 489 ->send_arg ("str3")
340 ->run ("Some::function", sub { 490 ->run ("Some::function", sub {
341 my ($fh) = @_; 491 my ($fh) = @_;
342 492
343 # fh is nonblocking, but we trust that the OS can accept these 493 # fh is nonblocking, but we trust that the OS can accept these
344 # extra 3 octets anyway. 494 # few octets anyway.
345 syswrite $fh, "hi #$_\n"; 495 syswrite $fh, "hi #$_\n";
346 496
347 # $fh is being closed here, as we don't store it anywhere 497 # $fh is being closed here, as we don't store it anywhere
348 }); 498 });
349 } 499 }
… …
351 # Some::function might look like this - all parameters passed before fork 501 # Some::function might look like this - all parameters passed before fork
352 # and after will be passed, in order, after the communications socket. 502 # and after will be passed, in order, after the communications socket.
353 sub Some::function { 503 sub Some::function {
354 my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_; 504 my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_;
355 505
356 print scalar <$fh>; # prints "hi 1\n" and "hi 2\n" 506 print scalar <$fh>; # prints "hi #1\n" and "hi #2\n" in any order
357 } 507 }
508
509PERFORMANCE
510 Now for some unscientific benchmark numbers (all done on an amd64
511 GNU/Linux box). These are intended to give you an idea of the relative
512 performance you can expect, they are not meant to be absolute
513 performance numbers.
514
515 OK, so, I ran a simple benchmark that creates a socket pair, forks,
516 calls exit in the child and waits for the socket to close in the parent.
517 I did load AnyEvent, EV and AnyEvent::Fork, for a total process size of
518 5100kB.
519
520 2079 new processes per second, using manual socketpair + fork
521
522 Then I did the same thing, but instead of calling fork, I called
523 AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the
524 socket form the child to close on exit. This does the same thing as
525 manual socket pair + fork, except that what is forked is the template
526 process (2440kB), and the socket needs to be passed to the server at the
527 other end of the socket first.
528
529 2307 new processes per second, using AnyEvent::Fork->new
530
531 And finally, using "new_exec" instead "new", using vforks+execs to exec
532 a new perl interpreter and compile the small server each time, I get:
533
534 479 vfork+execs per second, using AnyEvent::Fork->new_exec
535
536 So how can "AnyEvent->new" be faster than a standard fork, even though
537 it uses the same operations, but adds a lot of overhead?
538
539 The difference is simply the process size: forking the 5MB process takes
540 so much longer than forking the 2.5MB template process that the extra
541 overhead is canceled out.
542
543 If the benchmark process grows, the normal fork becomes even slower:
544
545 1340 new processes, manual fork of a 20MB process
546 731 new processes, manual fork of a 200MB process
547 235 new processes, manual fork of a 2000MB process
548
549 What that means (to me) is that I can use this module without having a
550 bad conscience because of the extra overhead required to start new
551 processes.
358 552
359TYPICAL PROBLEMS 553TYPICAL PROBLEMS
360 This section lists typical problems that remain. I hope by recognising 554 This section lists typical problems that remain. I hope by recognising
361 them, most can be avoided. 555 them, most can be avoided.
362 556
363 "leaked" file descriptors for exec'ed processes 557 leaked file descriptors for exec'ed processes
364 POSIX systems inherit file descriptors by default when exec'ing a 558 POSIX systems inherit file descriptors by default when exec'ing a
365 new process. While perl itself laudably sets the close-on-exec flags 559 new process. While perl itself laudably sets the close-on-exec flags
366 on new file handles, most C libraries don't care, and even if all 560 on new file handles, most C libraries don't care, and even if all
367 cared, it's often not possible to set the flag in a race-free 561 cared, it's often not possible to set the flag in a race-free
368 manner. 562 manner.
369 563
370 That means some file descriptors can leak through. And since it 564 That means some file descriptors can leak through. And since it
371 isn't possible to know which file descriptors are "good" and 565 isn't possible to know which file descriptors are "good" and
372 "neccessary" (or even to know which file descreiptors are open), 566 "necessary" (or even to know which file descriptors are open), there
373 there is no good way to close the ones that might harm. 567 is no good way to close the ones that might harm.
374 568
375 As an example of what "harm" can be done consider a web server that 569 As an example of what "harm" can be done consider a web server that
376 accepts connections and afterwards some module uses AnyEvent::Fork 570 accepts connections and afterwards some module uses AnyEvent::Fork
377 for the first time, causing it to fork and exec a new process, which 571 for the first time, causing it to fork and exec a new process, which
378 might inherit the network socket. When the server closes the socket, 572 might inherit the network socket. When the server closes the socket,
… …
385 exec'ed well before many random file descriptors are open. 579 exec'ed well before many random file descriptors are open.
386 580
387 In general, the solution for these kind of problems is to fix the 581 In general, the solution for these kind of problems is to fix the
388 libraries or the code that leaks those file descriptors. 582 libraries or the code that leaks those file descriptors.
389 583
390 Fortunately, most of these lekaed descriptors do no harm, other than 584 Fortunately, most of these leaked descriptors do no harm, other than
391 sitting on some resources. 585 sitting on some resources.
392 586
393 "leaked" file descriptors for fork'ed processes 587 leaked file descriptors for fork'ed processes
394 Normally, AnyEvent::Fork does start new processes by exec'ing them, 588 Normally, AnyEvent::Fork does start new processes by exec'ing them,
395 which closes file descriptors not marked for being inherited. 589 which closes file descriptors not marked for being inherited.
396 590
397 However, AnyEvent::Fork::Early and AnyEvent::Fork::Template offer a 591 However, AnyEvent::Fork::Early and AnyEvent::Fork::Template offer a
398 way to create these processes by forking, and this leaks more file 592 way to create these processes by forking, and this leaks more file
… …
405 trouble with a fork. 599 trouble with a fork.
406 600
407 The solution is to either not load these modules before use'ing 601 The solution is to either not load these modules before use'ing
408 AnyEvent::Fork::Early or AnyEvent::Fork::Template, or to delay 602 AnyEvent::Fork::Early or AnyEvent::Fork::Template, or to delay
409 initialising them, for example, by calling "init Gtk2" manually. 603 initialising them, for example, by calling "init Gtk2" manually.
604
605 exiting calls object destructors
606 This only applies to users of AnyEvent::Fork:Early and
607 AnyEvent::Fork::Template, or when initialising code creates objects
608 that reference external resources.
609
610 When a process created by AnyEvent::Fork exits, it might do so by
611 calling exit, or simply letting perl reach the end of the program.
612 At which point Perl runs all destructors.
613
614 Not all destructors are fork-safe - for example, an object that
615 represents the connection to an X display might tell the X server to
616 free resources, which is inconvenient when the "real" object in the
617 parent still needs to use them.
618
619 This is obviously not a problem for AnyEvent::Fork::Early, as you
620 used it as the very first thing, right?
621
622 It is a problem for AnyEvent::Fork::Template though - and the
623 solution is to not create objects with nontrivial destructors that
624 might have an effect outside of Perl.
410 625
411PORTABILITY NOTES 626PORTABILITY NOTES
412 Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a 627 Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a
413 nop, and ::Template is not going to work), and it cost a lot of blood 628 nop, and ::Template is not going to work), and it cost a lot of blood
414 and sweat to make it so, mostly due to the bloody broken perl that 629 and sweat to make it so, mostly due to the bloody broken perl that
415 nobody seems to care about. The fork emulation is a bad joke - I have 630 nobody seems to care about. The fork emulation is a bad joke - I have
416 yet to see something useful that you cna do with it without running into 631 yet to see something useful that you can do with it without running into
417 memory corruption issues or other braindamage. Hrrrr. 632 memory corruption issues or other braindamage. Hrrrr.
418 633
419 Cygwin perl is not supported at the moment, as it should implement fd 634 Cygwin perl is not supported at the moment due to some hilarious
420 passing, but doesn't, and rolling my own is hard, as cygwin doesn't 635 shortcomings of its API - see IO::FDPoll for more details.
421 support enough functionality to do it.
422 636
423SEE ALSO 637SEE ALSO
424 AnyEvent::Fork::Early (to avoid executing a perl interpreter), 638 AnyEvent::Fork::Early, to avoid executing a perl interpreter at all
639 (part of this distribution).
640
425 AnyEvent::Fork::Template (to create a process by forking the main 641 AnyEvent::Fork::Template, to create a process by forking the main
426 program at a convenient time). 642 program at a convenient time (part of this distribution).
427 643
428AUTHOR 644 AnyEvent::Fork::RPC, for simple RPC to child processes (on CPAN).
645
646AUTHOR AND CONTACT INFORMATION
429 Marc Lehmann <schmorp@schmorp.de> 647 Marc Lehmann <schmorp@schmorp.de>
430 http://home.schmorp.de/ 648 http://software.schmorp.de/pkg/AnyEvent-Fork
431 649

Diff Legend

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