NAME¶

Net::Server::PreForkSimple - Net::Server personality

SYNOPSIS¶

    use base qw(Net::Server::PreForkSimple);
    sub process_request {
        #...code...
    }
    __PACKAGE__->run();

DESCRIPTION¶

Please read the pod on Net::Server first. This module is a personality, or extension, or sub class, of the Net::Server module.

This personality binds to one or more ports and then forks "max_servers" child processes. The server will make sure that at any given time there are always "max_servers" available to receive a client request. Each of these children will process up to "max_requests" client connections. This type is good for a heavily hit site that can keep "max_servers" processes dedicated to the serving. (Multi port accept defaults to using flock to serialize the children).

At this time, it does not appear that this module will pass tests on Win32 systems. Any ideas or patches for making the tests pass would be welcome.

SAMPLE CODE¶

Please see the sample listed in Net::Server.

COMMAND LINE ARGUMENTS¶

In addition to the command line arguments of the Net::Server base class, Net::Server::PreFork contains several other configurable parameters.

    Key               Value                   Default
    max_servers       \d+                     50
    max_requests      \d+                     1000
    serialize         (flock|semaphore
                       |pipe|none)  undef
    # serialize defaults to flock on multi_port or on Solaris
    lock_file         "filename"              File::Temp->new
    check_for_dead    \d+                     30
    max_dequeue       \d+                     undef
    check_for_dequeue \d+                     undef

max_servers: The maximum number of child servers to start and maintain. This does not apply to dequeue processes.
max_requests: The number of client connections to receive before a child terminates.
serialize: Determines whether the server serializes child connections. Options are undef, flock, semaphore, pipe, or none. Default is undef. On multi_port servers or on servers running on Solaris, the default is flock. The flock option uses blocking exclusive flock on the file specified in lock_file (see below). The semaphore option uses IPC::Semaphore (thanks to Bennett Todd) for giving some sample code. The pipe option reads on a pipe to choose the next. the flock option should be the most bulletproof while the pipe option should be the most portable. (Flock is able to relinquish the block if the process dies between accept on the socket and reading of the client connection - semaphore and pipe do not). An option of none will not perform any serialization. If "none" is passed and there are multiple ports then a the default serialization will be used instead of "none."
lock_file: Filename to use in flock serialized accept in order to serialize the accept sequence between the children. This will default to a generated temporary filename. If default value is used the lock_file will be removed when the server closes.
check_for_dead: Seconds to wait before checking to see if a child died without letting the parent know.
max_dequeue: The maximum number of dequeue processes to start. If a value of zero or undef is given, no dequeue processes will be started. The number of running dequeue processes will be checked by the check_for_dead variable.
check_for_dequeue: Seconds to wait before forking off a dequeue process. The run_dequeue hook must be defined when using this setting. It is intended to use the dequeue process to take care of items such as mail queues. If a value of undef is given, no dequeue processes will be started.

CONFIGURATION FILE¶

"Net::Server::PreFork" allows for the use of a configuration file to read in server parameters. The format of this conf file is simple key value pairs. Comments and white space are ignored.

    #-------------- file test.conf --------------
    ### server information
    max_servers   80
    max_requests  1000
    ### user and group to become
    user        somebody
    group       everybody
    ### logging ?
    log_file    /var/log/server.log
    log_level   3
    pid_file    /tmp/server.pid
    ### access control
    allow       .+\.(net|com)
    allow       domain\.com
    deny        a.+
    ### background the process?
    background  1
    ### ports to bind
    host        127.0.0.1
    port        localhost:20204
    port        20205
    ### reverse lookups ?
    # reverse_lookups on
    #-------------- file test.conf --------------

PROCESS FLOW¶

Process flow follows Net::Server until the loop phase. At this point "max_servers" are forked and wait for connections. When a child accepts a connection, finishes processing a client, or exits, it relays that information to the parent, which keeps track and makes sure there are always "max_servers" running.

HOOKS¶

The PreForkSimple server has the following hooks in addition to the hooks provided by the Net::Server base class. See Net::Server

"$self->run_n_children_hook()": This hook occurs at the top of run_n_children which is called each time the server goes to start more child processes. This gives the parent to do a little of its own accounting (as desired). Idea for this hook came from James FitzGibbon.
"$self->child_init_hook()": This hook takes place immediately after the child process forks from the parent and before the child begins accepting connections. It is intended for any additional chrooting or other security measures. It is suggested that all perl modules be used by this point, so that the most shared memory possible is used.
"$self->child_finish_hook()": This hook takes place immediately before the child tells the parent that it is exiting. It is intended for saving out logged information or other general cleanup.
"$self->run_dequeue()": This hook only gets called in conjunction with the check_for_dequeue setting.
"$self->idle_loop_hook()": This hook is called in every pass through the main process wait loop.

HOT DEPLOY¶

Since version 2.000, the PreForkSimple server has accepted the TTIN and TTOU signals. When a TTIN is received, the max_servers is increased by 1. If a TTOU signal is received the max_servers is decreased by 1. This allows for adjusting the number of handling processes without having to restart the server.

BUGS¶

Tests don't seem to work on Win32. Any ideas or patches would be welcome.

TO DO¶

See Net::Server

AUTHOR¶

Paul T. Seamons paul@seamons.com

THANKS¶