AnyEvent::Fork - everything you wanted to use fork() for, but couldn't


   use AnyEvent::Fork;

      ->require ("MyModule")
      ->run ("MyModule::server", my $cv = AE::cv);

   my $fh = $cv->recv;


This module allows you to create new processes, without actually forking them from your current process (avoiding the problems of forking), but preserving most of the advantages of fork.

It can be used to create new worker processes or new independent subprocesses for short- and long-running jobs, process pools (e.g. for use in pre-forked servers) but also to spawn new external processes (such as CGI scripts from a web server), which can be faster (and more well behaved) than using fork+exec in big processes.

Special care has been taken to make this module useful from other modules, while still supporting specialised environments such as App::Staticperl or PAR::Packer.


This module only creates processes and lets you pass file handles and strings to it, and run perl code. It does not implement any kind of RPC - there is no back channel from the process back to you, and there is no RPC or message passing going on.

If you need some form of RPC, you could use the AnyEvent::Fork::RPC companion module, which adds simple RPC/job queueing to a process created by this module.

And if you need some automatic process pool management on top of AnyEvent::Fork::RPC, you can look at the AnyEvent::Fork::Pool companion module.

Or you can implement it yourself in whatever way you like: use some message-passing module such as AnyEvent::MP, some pipe such as AnyEvent::ZeroMQ, use AnyEvent::Handle on both sides to send e.g. JSON or Storable messages, and so on.


There is an abundance of modules on CPAN that do "something fork", such as Parallel::ForkManager, AnyEvent::ForkManager, AnyEvent::Worker or AnyEvent::Subprocess. There are modules that implement their own process management, such as AnyEvent::DBI.

The problems that all these modules try to solve are real, however, none of them (from what I have seen) tackle the very real problems of unwanted memory sharing, efficiency or not being able to use event processing, GUI toolkits or similar modules in the processes they create.

This module doesn't try to replace any of them - instead it tries to solve the problem of creating processes with a minimum of fuss and overhead (and also luxury). Ideally, most of these would use AnyEvent::Fork internally, except they were written before AnyEvent:Fork was available, so obviously had to roll their own.


There are two traditional ways to implement parallel processing on UNIX like operating systems - fork and process, and fork+exec and process. They have different advantages and disadvantages that I describe below, together with how this module tries to mitigate the disadvantages.

Forking from a big process can be very slow.

A 5GB process needs 0.05s to fork on my 3.6GHz amd64 GNU/Linux box. This overhead is often shared with exec (because you have to fork first), but in some circumstances (e.g. when vfork is used), fork+exec can be much faster.

This module can help here by telling a small(er) helper process to fork, which is faster then forking the main process, and also uses vfork where possible. This gives the speed of vfork, with the flexibility of fork.

Forking usually creates a copy-on-write copy of the parent process.

For example, modules or data files that are loaded will not use additional memory after a fork. Exec'ing a new process, in contrast, means modules and data files might need to be loaded again, at extra CPU and memory cost.

But when forking, you still create a copy of your data structures - if the program frees them and replaces them by new data, the child processes will retain the old version even if it isn't used, which can suddenly and unexpectedly increase memory usage when freeing memory.

For example, Gtk2::CV is an image viewer optimised for large directories (millions of pictures). It also forks subprocesses for thumbnail generation, which inherit the data structure that stores all file information. If the user changes the directory, it gets freed in the main process, leaving a copy in the thumbnailer processes. This can lead to many times the memory usage that would actually be required. The solution is to fork early (and being unable to dynamically generate more subprocesses or do this from a module)... or to use AnyEvent:Fork.

There is a trade-off between more sharing with fork (which can be good or bad), and no sharing with exec.

This module allows the main program to do a controlled fork, and allows modules to exec processes safely at any time. When creating a custom process pool you can take advantage of data sharing via fork without risking to share large dynamic data structures that will blow up child memory usage.

In other words, this module puts you into control over what is being shared and what isn't, at all times.

Exec'ing a new perl process might be difficult.

For example, it is not easy to find the correct path to the perl interpreter - $^X might not be a perl interpreter at all. Worse, there might not even be a perl binary installed on the system.

This module tries hard to identify the correct path to the perl interpreter. With a cooperative main program, exec'ing the interpreter might not even be necessary, but even without help from the main program, it will still work when used from a module.

Exec'ing a new perl process might be slow, as all necessary modules have to be loaded from disk again, with no guarantees of success.

Long running processes might run into problems when perl is upgraded and modules are no longer loadable because they refer to a different perl version, or parts of a distribution are newer than the ones already loaded.

This module supports creating pre-initialised perl processes to be used as a template for new processes at a later time, e.g. for use in a process pool.

Forking might be impossible when a program is running.

For example, POSIX makes it almost impossible to fork from a multi-threaded program while doing anything useful in the child - in fact, if your perl program uses POSIX threads (even indirectly via e.g. IO::AIO or threads), you cannot call fork on the perl level anymore without risking memory corruption or worse on a number of operating systems.

This module can safely fork helper processes at any time, by calling fork+exec in C, in a POSIX-compatible way (via Proc::FastSpawn).

Parallel processing with fork might be inconvenient or difficult to implement. Modules might not work in both parent and child.

For example, when a program uses an event loop and creates watchers it becomes very hard to use the event loop from a child program, as the watchers already exist but are only meaningful in the parent. Worse, a module might want to use such a module, not knowing whether another module or the main program also does, leading to problems.

Apart from event loops, graphical toolkits also commonly fall into the "unsafe module" category, or just about anything that communicates with the external world, such as network libraries and file I/O modules, which usually don't like being copied and then allowed to continue in two processes.

With this module only the main program is allowed to create new processes by forking (because only the main program can know when it is still safe to do so) - all other processes are created via fork+exec, which makes it possible to use modules such as event loops or window interfaces safely.


This is where the wall of text ends and code speaks.

Create a single new process, tell it to run your worker function.

      ->require ("MyModule")
      ->run ("MyModule::worker, sub {
         my ($master_filehandle) = @_;

         # now $master_filehandle is connected to the
         # $slave_filehandle in the new process.

MyModule might look like this:

   package MyModule;

   sub worker {
      my ($slave_filehandle) = @_;

      # now $slave_filehandle is connected to the $master_filehandle
      # in the original process. have fun!

Create a pool of server processes all accepting on the same socket.

   # create listener socket
   my $listener = ...;

   # create a pool template, initialise it and give it the socket
   my $pool = AnyEvent::Fork
                 ->require ("Some::Stuff", "My::Server")
                 ->send_fh ($listener);

   # now create 10 identical workers
   for my $id (1..10) {
         ->send_arg ($id)
         ->run ("My::Server::run");

   # now do other things - maybe use the filehandle provided by run
   # to wait for the processes to die. or whatever.

My::Server might look like this:

   package My::Server;

   sub run {
      my ($slave, $listener, $id) = @_;

      close $slave; # we do not use the socket, so close it to save resources

      # we could go ballistic and use e.g. AnyEvent here, or IO::AIO,
      # or anything we usually couldn't do in a process forked normally.
      while (my $socket = $listener->accept) {
         # do sth. with new socket

use AnyEvent::Fork as a faster fork+exec

This runs /bin/echo hi, with standard output redirected to /tmp/log and standard error redirected to the communications socket. It is usually faster than fork+exec, but still lets you prepare the environment.

   open my $output, ">/tmp/log" or die "$!";

      ->eval ('
           # compile a helper function for later use
           sub run {
              my ($fh, $output, @cmd) = @_;

              # perl will clear close-on-exec on STDOUT/STDERR
              open STDOUT, ">&", $output or die;
              open STDERR, ">&", $fh or die;

              exec @cmd;
      ->send_fh ($output)
      ->send_arg ("/bin/echo", "hi")
      ->run ("run", my $cv = AE::cv);

   my $stderr = $cv->recv;

For stingy users: put the worker code into a DATA section.

When you want to be stingy with files, you can put your code into the DATA section of your module (or program):

   use AnyEvent::Fork;

      ->eval (do { local $/; <DATA> })
      ->run ("doit", sub { ... });


   sub doit {
      ... do something!

For stingy standalone programs: do not rely on external files at all.

For single-file scripts it can be inconvenient to rely on external files - even when using a DATA section, you still need to exec an external perl interpreter, which might not be available when using App::Staticperl, Urlader or PAR::Packer for example.

Two modules help here - AnyEvent::Fork::Early forks a template process for all further calls to new_exec, and AnyEvent::Fork::Template forks the main program as a template process.

Here is how your main program should look like:

   #! perl

   # optional, as the very first thing.
   # in case modules want to create their own processes.
   use AnyEvent::Fork::Early;

   # next, load all modules you need in your template process
   use Example::My::Module
   use Example::Whatever;

   # next, put your run function definition and anything else you
   # need, but do not use code outside of BEGIN blocks.
   sub worker_run {
      my ($fh, @args) = @_;

   # now preserve everything so far as AnyEvent::Fork object
   # in $TEMPLATE.
   use AnyEvent::Fork::Template;

   # do not put code outside of BEGIN blocks until here

   # now use the $TEMPLATE process in any way you like

   # for example: create 10 worker processes
   my @worker;
   my $cv = AE::cv;
   for (1..10) {
      $TEMPLATE->fork->send_arg ($_)->run ("worker_run", sub {
         push @worker, shift;


This module can create new processes either by executing a new perl process, or by forking from an existing "template" process.

All these processes are called "child processes" (whether they are direct children or not), while the process that manages them is called the "parent process".

Each such process comes with its own file handle that can be used to communicate with it (it's actually a socket - one end in the new process, one end in the main process), and among the things you can do in it are load modules, fork new processes, send file handles to it, and execute functions.

There are multiple ways to create additional processes to execute some jobs:

fork a new process from the "default" template process, load code, run it

This module has a "default" template process which it executes when it is needed the first time. Forking from this process shares the memory used for the perl interpreter with the new process, but loading modules takes time, and the memory is not shared with anything else.

This is ideal for when you only need one extra process of a kind, with the option of starting and stopping it on demand.


      ->require ("Some::Module")
      ->run ("Some::Module::run", sub {
         my ($fork_fh) = @_;
fork a new template process, load code, then fork processes off of it and run the code

When you need to have a bunch of processes that all execute the same (or very similar) tasks, then a good way is to create a new template process for them, loading all the modules you need, and then create your worker processes from this new template process.

This way, all code (and data structures) that can be shared (e.g. the modules you loaded) is shared between the processes, and each new process consumes relatively little memory of its own.

The disadvantage of this approach is that you need to create a template process for the sole purpose of forking new processes from it, but if you only need a fixed number of processes you can create them, and then destroy the template process.


   my $template = AnyEvent::Fork->new->require ("Some::Module");
   for (1..10) {
      $template->fork->run ("Some::Module::run", sub {
         my ($fork_fh) = @_;

   # at this point, you can keep $template around to fork new processes
   # later, or you can destroy it, which causes it to vanish.
execute a new perl interpreter, load some code, run it

This is relatively slow, and doesn't allow you to share memory between multiple processes.

The only advantage is that you don't have to have a template process hanging around all the time to fork off some new processes, which might be an advantage when there are long time spans where no extra processes are needed.


      ->require ("Some::Module")
      ->run ("Some::Module::run", sub {
         my ($fork_fh) = @_;

THE AnyEvent::Fork CLASS

This module exports nothing, and only implements a single class - AnyEvent::Fork.

There are two class constructors that both create new processes - new and new_exec. The fork method creates a new process by forking an existing one and could be considered a third constructor.

Most of the remaining methods deal with preparing the new process, by loading code, evaluating code and sending data to the new process. They usually return the process object, so you can chain method calls.

If a process object is destroyed before calling its run method, then the process simply exits. After run is called, all responsibility is passed to the specified function.

As long as there is any outstanding work to be done, process objects resist being destroyed, so there is no reason to store them unless you need them later - configure and forget works just fine.

my $proc = new AnyEvent::Fork

Create a new "empty" perl interpreter process and returns its process object for further manipulation.

The new process is forked from a template process that is kept around for this purpose. When it doesn't exist yet, it is created by a call to new_exec first and then stays around for future calls.

$new_proc = $proc->fork

Forks $proc, creating a new process, and returns the process object of the new process.

If any of the send_ functions have been called before fork, then they will be cloned in the child. For example, in a pre-forked server, you might send_fh the listening socket into the template process, and then keep calling fork and run.

my $proc = new_exec AnyEvent::Fork

Create a new "empty" perl interpreter process and returns its process object for further manipulation.

Unlike the new method, this method always spawns a new perl process (except in some cases, see AnyEvent::Fork::Early for details). This reduces the amount of memory sharing that is possible, and is also slower.

You should use new whenever possible, except when having a template process around is unacceptable.

The path to the perl interpreter is divined using various methods - first $^X is investigated to see if the path ends with something that looks as if it were the perl interpreter. Failing this, the module falls back to using $Config::Config{perlpath}.

The path to perl can also be overridden by setting the global variable $AnyEvent::Fork::PERL - it's value will be used for all subsequent invocations.

$pid = $proc->pid

Returns the process id of the process iff it is a direct child of the process running AnyEvent::Fork, and undef otherwise. As a general rule (that you cannot rely upon), processes created via new_exec, AnyEvent::Fork::Early or AnyEvent::Fork::Template are direct children, while all other processes are not.

Or in other words, you do not normally have to take care of zombies for processes created via new, but when in doubt, or zombies are a problem, you need to check whether a process is a diretc child by calling this method, and possibly creating a child watcher or reap it manually.

$proc = $proc->eval ($perlcode, @args)

Evaluates the given $perlcode as ... Perl code, while setting @_ to the strings specified by @args, in the "main" package (so you can access the args using $_[0] and so on, but not using implicit shit as the latter works on @ARGV).

This call is meant to do any custom initialisation that might be required (for example, the require method uses it). It's not supposed to be used to completely take over the process, use run for that.

The code will usually be executed after this call returns, and there is no way to pass anything back to the calling process. Any evaluation errors will be reported to stderr and cause the process to exit.

If you want to execute some code (that isn't in a module) to take over the process, you should compile a function via eval first, and then call it via run. This also gives you access to any arguments passed via the send_xxx methods, such as file handles. See the "use AnyEvent::Fork as a faster fork+exec" example to see it in action.

Returns the process object for easy chaining of method calls.

It's common to want to call an iniitalisation function with some arguments. Make sure you actually pass @_ to that function (for example by using &name syntax), and do not just specify a function name:

   $proc->eval ('&MyModule::init', $string1, $string2);
$proc = $proc->require ($module, ...)

Tries to load the given module(s) into the process

Returns the process object for easy chaining of method calls.

$proc = $proc->send_fh ($handle, ...)

Send one or more file handles (not file descriptors) to the process, to prepare a call to run.

The process object keeps a reference to the handles until they have been passed over to the process, so you must not explicitly close the handles. This is most easily accomplished by simply not storing the file handles anywhere after passing them to this method - when AnyEvent::Fork is finished using them, perl will automatically close them.

Returns the process object for easy chaining of method calls.

Example: pass a file handle to a process, and release it without closing. It will be closed automatically when it is no longer used.

   $proc->send_fh ($my_fh);
   undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT
$proc = $proc->send_arg ($string, ...)

Send one or more argument strings to the process, to prepare a call to run. The strings can be any octet strings.

The protocol is optimised to pass a moderate number of relatively short strings - while you can pass up to 4GB of data in one go, this is more meant to pass some ID information or other startup info, not big chunks of data.

Returns the process object for easy chaining of method calls.

$proc->run ($func, $cb->($fh))

Enter the function specified by the function name in $func in the process. The function is called with the communication socket as first argument, followed by all file handles and string arguments sent earlier via send_fh and send_arg methods, in the order they were called.

The process object becomes unusable on return from this function - any further method calls result in undefined behaviour.

The function name should be fully qualified, but if it isn't, it will be looked up in the main package.

If the called function returns, doesn't exist, or any error occurs, the process exits.

Preparing the process is done in the background - when all commands have been sent, the callback is invoked with the local communications socket as argument. At this point you can start using the socket in any way you like.

If the communication socket isn't used, it should be closed on both sides, to save on kernel memory.

The socket is non-blocking in the parent, and blocking in the newly created process. The close-on-exec flag is set in both.

Even if not used otherwise, the socket can be a good indicator for the existence of the process - if the other process exits, you get a readable event on it, because exiting the process closes the socket (if it didn't create any children using fork).

Compatibility to AnyEvent::Fork::Remote

If you want to write code that works with both this module and AnyEvent::Fork::Remote, you need to write your code so that it assumes there are two file handles for communications, which might not be unix domain sockets. The run function should start like this:

   sub run {
      my ($rfh, @args) = @_; # @args is your normal arguments
      my $wfh = fileno $rfh ? $rfh : *STDOUT;

      # now use $rfh for reading and $wfh for writing

This checks whether the passed file handle is, in fact, the process STDIN handle. If it is, then the function was invoked visa AnyEvent::Fork::Remote, so STDIN should be used for reading and STDOUT should be used for writing.

In all other cases, the function was called via this module, and there is only one file handle that should be sued for reading and writing.

Example: create a template for a process pool, pass a few strings, some file handles, then fork, pass one more string, and run some code.

   my $pool = AnyEvent::Fork
                 ->send_arg ("str1", "str2")
                 ->send_fh ($fh1, $fh2);

   for (1..2) {
         ->send_arg ("str3")
         ->run ("Some::function", sub {
            my ($fh) = @_;

            # fh is nonblocking, but we trust that the OS can accept these
            # few octets anyway.
            syswrite $fh, "hi #$_\n";

            # $fh is being closed here, as we don't store it anywhere

   # Some::function might look like this - all parameters passed before fork
   # and after will be passed, in order, after the communications socket.
   sub Some::function {
      my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_;

      print scalar <$fh>; # prints "hi #1\n" and "hi #2\n" in any order


This module has a limited API for use in child processes.

@args = AnyEvent::Fork::Serve::run_args

This function, which only exists before the run method is called, returns the arguments that would be passed to the run function, and clears them.

This is mainly useful to get any file handles passed via send_fh, but works for any arguments passed via send_xxx methods.


These methods might go away completely or change behaviour, at any time.

$proc->to_fh ($cb->($fh)) # EXPERIMENTAL, MIGHT BE REMOVED

Flushes all commands out to the process and then calls the callback with the communications socket.

The process object becomes unusable on return from this function - any further method calls result in undefined behaviour.

The point of this method is to give you a file handle that you can pass to another process. In that other process, you can call new_from_fh AnyEvent::Fork $fh to create a new AnyEvent::Fork object from it, thereby effectively passing a fork object to another process.

new_from_fh AnyEvent::Fork $fh # EXPERIMENTAL, MIGHT BE REMOVED

Takes a file handle originally rceeived by the to_fh method and creates a new AnyEvent:Fork object. The child process itself will not change in any way, i.e. it will keep all the modifications done to it before calling to_fh.

The new object is very much like the original object, except that the pid method will return undef even if the process is a direct child.


Now for some unscientific benchmark numbers (all done on an amd64 GNU/Linux box). These are intended to give you an idea of the relative performance you can expect, they are not meant to be absolute performance numbers.

OK, so, I ran a simple benchmark that creates a socket pair, forks, calls exit in the child and waits for the socket to close in the parent. I did load AnyEvent, EV and AnyEvent::Fork, for a total process size of 5100kB.

   2079 new processes per second, using manual socketpair + fork

Then I did the same thing, but instead of calling fork, I called AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the socket from the child to close on exit. This does the same thing as manual socket pair + fork, except that what is forked is the template process (2440kB), and the socket needs to be passed to the server at the other end of the socket first.

   2307 new processes per second, using AnyEvent::Fork->new

And finally, using new_exec instead new, using vforks+execs to exec a new perl interpreter and compile the small server each time, I get:

    479 vfork+execs per second, using AnyEvent::Fork->new_exec

So how can AnyEvent->new be faster than a standard fork, even though it uses the same operations, but adds a lot of overhead?

The difference is simply the process size: forking the 5MB process takes so much longer than forking the 2.5MB template process that the extra overhead is canceled out.

If the benchmark process grows, the normal fork becomes even slower:

   1340 new processes, manual fork of a 20MB process
    731 new processes, manual fork of a 200MB process
    235 new processes, manual fork of a 2000MB process

What that means (to me) is that I can use this module without having a bad conscience because of the extra overhead required to start new processes.


This section lists typical problems that remain. I hope by recognising them, most can be avoided.

leaked file descriptors for exec'ed processes

POSIX systems inherit file descriptors by default when exec'ing a new process. While perl itself laudably sets the close-on-exec flags on new file handles, most C libraries don't care, and even if all cared, it's often not possible to set the flag in a race-free manner.

That means some file descriptors can leak through. And since it isn't possible to know which file descriptors are "good" and "necessary" (or even to know which file descriptors are open), there is no good way to close the ones that might harm.

As an example of what "harm" can be done consider a web server that accepts connections and afterwards some module uses AnyEvent::Fork for the first time, causing it to fork and exec a new process, which might inherit the network socket. When the server closes the socket, it is still open in the child (which doesn't even know that) and the client might conclude that the connection is still fine.

For the main program, there are multiple remedies available - AnyEvent::Fork::Early is one, creating a process early and not using new_exec is another, as in both cases, the first process can be exec'ed well before many random file descriptors are open.

In general, the solution for these kind of problems is to fix the libraries or the code that leaks those file descriptors.

Fortunately, most of these leaked descriptors do no harm, other than sitting on some resources.

leaked file descriptors for fork'ed processes

Normally, AnyEvent::Fork does start new processes by exec'ing them, which closes file descriptors not marked for being inherited.

However, AnyEvent::Fork::Early and AnyEvent::Fork::Template offer a way to create these processes by forking, and this leaks more file descriptors than exec'ing them, as there is no way to mark descriptors as "close on fork".

An example would be modules like EV, IO::AIO or Gtk2. Both create pipes for internal uses, and Gtk2 might open a connection to the X server. EV and IO::AIO can deal with fork, but Gtk2 might have trouble with a fork.

The solution is to either not load these modules before use'ing AnyEvent::Fork::Early or AnyEvent::Fork::Template, or to delay initialising them, for example, by calling init Gtk2 manually.

exiting calls object destructors

This only applies to users of AnyEvent::Fork:Early and AnyEvent::Fork::Template, or when initialising code creates objects that reference external resources.

When a process created by AnyEvent::Fork exits, it might do so by calling exit, or simply letting perl reach the end of the program. At which point Perl runs all destructors.

Not all destructors are fork-safe - for example, an object that represents the connection to an X display might tell the X server to free resources, which is inconvenient when the "real" object in the parent still needs to use them.

This is obviously not a problem for AnyEvent::Fork::Early, as you used it as the very first thing, right?

It is a problem for AnyEvent::Fork::Template though - and the solution is to not create objects with nontrivial destructors that might have an effect outside of Perl.


Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop, and ::Template is not going to work), and it cost a lot of blood and sweat to make it so, mostly due to the bloody broken perl that nobody seems to care about. The fork emulation is a bad joke - I have yet to see something useful that you can do with it without running into memory corruption issues or other braindamage. Hrrrr.

Since fork is endlessly broken on win32 perls (it doesn't even remotely work within it's documented limits) and quite obviously it's not getting improved any time soon, the best way to proceed on windows would be to always use new_exec and thus never rely on perl's fork "emulation".

Cygwin perl is not supported at the moment due to some hilarious shortcomings of its API - see IO::FDPoll for more details. If you never use send_fh and always use new_exec to create processes, it should work though.


AnyEvent::Fork itself cannot generally be used in subprocesses. As long as only one process ever forks new processes, sharing the template processes is possible (you could use a pipe as a lock by writing a byte into it to unlock, and reading the byte to lock for example)

To make concurrent calls possible after fork, you should get rid of the template and early fork processes. AnyEvent::Fork will create a new template process as needed.

   undef $AnyEvent::Fork::EARLY;
   undef $AnyEvent::Fork::TEMPLATE;

It doesn't matter whether you get rid of them in the parent or child after a fork.


AnyEvent::Fork::Early, to avoid executing a perl interpreter at all (part of this distribution).

AnyEvent::Fork::Template, to create a process by forking the main program at a convenient time (part of this distribution).

AnyEvent::Fork::Remote, for another way to create processes that is mostly compatible to this module and modules building on top of it, but works better with remote processes.

AnyEvent::Fork::RPC, for simple RPC to child processes (on CPAN).

AnyEvent::Fork::Pool, for simple worker process pool (on CPAN).


 Marc Lehmann <>