Provided by: speedy-cgi-perl_2.22-13build2_amd64 bug


       SpeedyCGI - Speed up perl scripts by running them persistently.



        ### Your Script Here.  For example:
        print "Content-type: text/html\n\nHello World!\n";

        ## Optionally, use the CGI::SpeedyCGI module for various things

        # Create a SpeedyCGI object
        use CGI::SpeedyCGI;
        my $sp = CGI::SpeedyCGI->new;

        # See if we are running under SpeedyCGI or not.
        print "Running under speedy=", $sp->i_am_speedy ? 'yes' : 'no', "\n";

        # Register a shutdown handler
        $sp->add_shutdown_handler(sub { do something here });

        # Register a cleanup handler
        $sp->register_cleanup(sub { do something here });

        # Set/get some SpeedyCGI options
        $sp->setopt('timeout', 30);
        print "maxruns=", $sp->getopt('maxruns'), "\n";


       SpeedyCGI is a way to run perl scripts persistently, which can make them run much more
       quickly.  A script can be made to to run persistently by changing the interpreter line at
       the top of the script from:




       After the script is initially run, instead of exiting, the perl interpreter is kept
       running.  During subsequent runs, this interpreter is used to handle new executions
       instead of starting a new perl interpreter each time.  A very fast frontend program,
       written in C, is executed for each request.  This fast frontend then contacts the
       persistent Perl process, which is usually already running, to do the work and return the

       By default each perl script runs in its own Unix process, so one perl script can't
       interfere with another.  Command line options can also be used to deal with programs that
       have memory leaks or other problems that might keep them from otherwise running

       SpeedyCGI can be used to speed up perl CGI scripts.  It conforms to the CGI specification,
       and does not run perl code inside the web server.  Since the perl interpreter runs outside
       the web server, it can't cause problems for the web server itself.

       SpeedyCGI also provides an Apache module so that under the Apache web server, scripts can
       be run without the overhead of doing a fork/exec for each request.  With this module a
       small amount of frontend code is run within the web server - the perl interpreters still
       run outside the server.

       SpeedyCGI and PersistentPerl are currently both names for the same code.  SpeedyCGI was
       the original name, but because people weren't sure what it did, the name PersistentPerl
       was picked as an alias.  At some point SpeedyCGI will be replaced by PersistentPerl, or
       become a sub-class of PersistentPerl to avoid always having two distributions.


   Setting Option Values
       SpeedyCGI options can be set in several ways:

       Command Line
           The speedy command line is the same as for regular perl, with the exception that
           SpeedyCGI specific options can be passed in after a "--".

           For example the line:

                   #!/usr/bin/speedy -w -- -t300

           at the top of your script will set the perl option ""-w"" and will pass the ""-t""
           option to SpeedyCGI, setting the Timeout value to 300 seconds.

           Environment variables can be used to pass in options.  This can only be done before
           the initial execution, not from within the script itself.  The name of the environment
           variable is always SPEEDY_ followed by the option name in upper-case.  For example to
           set the speedy Timeout option, use the environment variable named SPEEDY_TIMEOUT.

           The CGI::SpeedyCGI module provides the setopt method to set options from within the
           perl script at runtime.  There is also a getopt method to retrieve the current
           options.  See "METHODS" below.

           If you are using the optional Apache module, SpeedyCGI options can be set in the
           httpd.conf file.  The name of the apache directive will always be Speedy followed by
           the option name.  For example to set the Timeout option, use the apache directive

           Note that these variables are global. There is currently no way to run different
           scripts with different SpeedyCGI options when they are run from the Apache module. Any
           <Directory> or <Location> contexts have no effect on the scope of the SpeedyCGI
           options. When the same SpeedyCGI option is set several times, the last one overrides
           the others.

       Not all options below are available in all contexts.  The context for which each option is
       valid is listed on the "Context" line in the section below.  There are three contexts:

           The command-line "speedy" program, used normally with #! at the top of your script or
           from a shell prompt.

           The optional Apache mod_speedycgi module.

           During perl execution via the CGI::SpeedyCGI module's getopt/setopt methods.

   Options Available
               Command Line    : -p<string>
               Default Value   : "/usr/bin/speedy_backend"
               Context         : mod_speedycgi, speedy


                   Path to the speedy backend program.

               Command Line    : -B<number>
               Default Value   : 131072
               Context         : speedy


                   Use <number> bytes as the maximum size for the buffer that
                   receives data from the perl backend.

               Command Line    : -b<number>
               Default Value   : 131072
               Context         : speedy


                   Use <number> bytes as the maximum size for the buffer that
                   sends data to the perl backend.

               Command Line    : -g<string>
               Default Value   : "none"
               Context         : mod_speedycgi, speedy


                   Allow a single perl interpreter to run multiple scripts.
                   All scripts that are run with the same group name and by
                   the same user will be run by the same group of perl
                   interpreters. If the group name is "none" then grouping is
                   disabled and each interpreter will run one script.
                   Different group names allow scripts to be separated into
                   different groups. Name is case-sensitive, and only the
                   first 12-characters are significant. Specifying an empty
                   group name is the same as specifying the group name
                   "default" - this allows just specifying "-g" on the command
                   line to turn on grouping.

               Command Line    : -M<number>
               Default Value   : 0 (no max)
               Context         : mod_speedycgi, speedy


                   If non-zero, limits the number of speedy backends running
                   for this perl script to <number>.

               Command Line    : -r<number>
               Default Value   : 500
               Context         : mod_speedycgi, module, speedy


                   Once the perl interpreter has run <number> times, re-exec
                   the backend process.  Zero indicates no maximum.  This
                   option is useful for processes that tend to consume
                   resources over time.

               Command Line    : N/A
               Default Value   : ""
               Context         : mod_speedycgi


                   Command-line options to pass to the perl interpreter.

               Command Line    : -t<number>
               Default Value   : 3600 (one hour)
               Context         : mod_speedycgi, module, speedy


                   If no new requests have been received after <number>
                   seconds, exit the persistent perl interpreter.  Zero
                   indicates no timeout.

               Command Line    : -T<string>
               Default Value   : "/tmp/speedy"
               Context         : mod_speedycgi, speedy


                   Use the given prefix for creating temporary files.  This
                   must be a filename prefix, not a directory name.

               Command Line    : -v
               Context         : speedy


                   Print the SpeedyCGI version and exit.


       The following methods are available in the CGI::SpeedyCGI module.

       new Create a new CGI::SpeedyCGI object.

               my $sp = CGI::SpeedyCGI->new;

           Register a function that will be called at the end of each request, after your script
           finishes running, but before STDOUT and STDERR are closed.  Multiple functions can be
           added by calling the method more than once.  At the end of the request, each function
           will be called in the order in which it was registered.


           Add a function to the list of functions that will be called right before the perl
           interpreter exits.  This is not at the end of each request, it is when the perl
           interpreter decides to exit completely due to a Timeout or reaching MaxRuns.

               $sp->add_shutdown_handler(sub {$dbh->logout});

           Deprecated.  Similar to "add_shutdown_handler", but only allows for a single function
           to be registered.

               $sp->set_shutdown_handler(sub {$dbh->logout});

           Returns a boolean telling whether this script is running under SpeedyCGI or not.  A
           perl script can run under regular perl, or under SpeedyCGI.  This method allows the
           script to tell which environment it is in.


           To make your script as portable as possible, you can use the following test to make
           sure both the SpeedyCGI module is available and you are running under SpeedyCGI:

               if (eval {require CGI::SpeedyCGI} && CGI::SpeedyCGI->i_am_speedy) {
                   Do something SpeedyCGI specific here...

           To increase the speed of this check you can also test whether the following variable
           is defined instead of going through the object interface:


       setopt($optname, $value)
           Set one of the SpeedyCGI options given in "Options Available".  Returns the option's
           previous value.  $optname is case-insensitive.

               $sp->setopt('TIMEOUT', 300);

           Return the current value of one of the SpeedyCGI options.  $optname is case-


           Shut down the perl interpreter right away.  This function does not return.


           Shut down the perl interpreter as soon as this request is done.



       To install SpeedyCGI you will need to either download a binary package for your OS, or
       compile SpeedyCGI from source code.  See "DOWNLOADING" for information on where to obtain
       the source code and binaries.

   Binary Installation
       Once you have downloaded the binary package for your OS, you'll need to install it using
       the normal package tools for your OS.  The commands to do that are:

            rpm -i <filename>

            gunzip <filename>.gz
            pkgadd -d <filename>

            pkg_add <filename>

       If you are also installing the apache module you will have to configure Apache as
       documented in "Apache Configuration".

   Source Code Installation
       To compile SpeedyCGI you will need perl 5.005_03 or later, and a C compiler, preferably
       the same one that your perl distribution was compiled with.  SpeedyCGI is known to work
       under Solaris, Redhat Linux, FreeBSD and OpenBSD.  There may be problems with other OSes
       or earlier versions of Perl.  SpeedyCGI may not work with threaded perl -- as of release
       2.10, Linux and Solaris seem to work OK with threaded perl, but FreeBSD does not.

   Standard Install
       To do a standard install from source code, execute the following:

           perl Makefile.PL
           make test
           make install

       This will install the speedy and speedy_backend binaries in the same directory where perl
       was installed, and the module in the standard perl lib directory.  It will
       also attempt to install the mod_speedycgi module if you have the command apxs in your

   Install in a Different Directory
       If you don't have permission to install into the standard perl directory, or if you want
       to install elsewhere, the easiest way is to compile and install your own copy of perl in
       another location, then use your new version of perl when you run "perl Makefile.PL".  The
       SpeedyCGI binaries and module will then be installed in the same location as the new
       version of perl.

       If you can't install your own perl, you can take the following steps:

       ·   Edit src/optdefs and change the default value for BackendProg to the location where
           speedy_backend will be installed.

       ·   Compile as above, then manually copy the speedy and speedy_backend binaries to where
           you want to install them.

       ·   If you want to use the CGI::SpeedyCGI module in your code (it's not required), you
           will have to use "use lib" so it can be located.

   Setuid Install
       SpeedyCGI has limited support for running setuid - installing this way may compromise the
       security of your system.  To install setuid do the following:

       ·   Run "perl Makefile.PL"

       ·   Edit speedy/Makefile and add "-DIAMSUID" to the end of the "DEFINE = " line.

       ·   Run make

       ·   Take the resulting "speedy" binary and install it suid-root as /usr/bin/speedy_suid

       ·   Change your setuid scripts to use /usr/bin/speedy_suid as the interpreter.

       This has been know to work in Linux and FreeBSD.  Solaris will work as long as the Group
       option is set to "none".

   Apache Installation
       To compile the optional apache mod_speedycgi module you must have the apxs command in your
       path.  Redhat includes this command with the "apache-devel" RPM, though it may not work
       properly for installation.

       If the apache installation fails:

       ·   Copy the from the mod_speedycgi directory, or from the
           mod_speedycgi2/.libs directory, to wherever your apache modules are stored (try

       ·   Edit your httpd.conf (try /etc/httpd/conf/httpd.conf) and add the following lines.
           The path at the end of the LoadModule directive may be different in your installation
           -- look at other LoadModules to see.

               LoadModule speedycgi_module modules/

           If you are using Apache-1, also add:

               AddModule mod_speedycgi.c

   Apache Configuration
       Once mod_speedycgi is installed, it has to be configured to be used for your perl scripts.
       There are two methods.

       Warning!  The instructions below may compromise the security of your web site.  The
       security risks associated with SpeedyCGI are similar to those of regular CGI.  If you
       don't understand the security implications of the changes below then don't make them.

       1. Path Configuration
           This is similar to the way /cgi-bin works - everything under this path is handled by
           SpeedyCGI.  Add the following lines near the top of your httpd.conf - this will cause
           all scripts in your cgi-bin directory to be handled by SpeedyCGI when they are
           accessed as /speedy/script-name.

               Alias /speedy/ /home/httpd/cgi-bin/
               <Location /speedy>
                   SetHandler speedycgi-script
                   Options ExecCGI
                   allow from all

       2. File Extension Configuration
           This will make SpeedyCGI handle all files with a certain extension, similar to the way
           .cgi files work.  Add the following lines near the top of your httpd.conf file - this
           will set up the file extension ".speedy" to be handled by SpeedyCGI.

               AddHandler speedycgi-script .speedy
               <Location />
                   Options ExecCGI


       How does the speedy front end connect to the back end process?
           Via a Unix socket in /tmp.  A queue is kept in a shared file in /tmp that holds an
           entry for each process.  In that queue are the pids of the perl processes waiting for
           connections.  The frontend pulls a process out of this queue, connects to its socket,
           sends over the environment and argv, and then uses this socket for stdin/stdout to the
           perl process.

       If another request comes in while SpeedyCGI script is running, does the client have to
       wait or is another process started?  Is there a way to set a limit on how many processes
       get started?
           If another request comes while all the perl processes are busy, then another perl
           process is started.  Just like in regular perl there is normally no limit on how many
           processes get started.  But, the processes are only started when the load is so high
           that they're necessary.  If the load goes down, the processes will die off due to
           inactivity, unless you disable the timeout.

           Starting in version 1.8.3 an option was added to limit the number of perl backends
           running.  See MaxBackends in "Options Available" above.

       How much of perl's state is kept when speedy starts another request? Do globals keep their
       values?  Are destructors run after the request?
           Globals keep their values.  Nothing is destroyed after the request.
           STDIN/STDOUT/STDERR are closed -- other files are not.  %ENV and @ARGV are the only
           globals changed between requests.

       How can I make sure speedy restarts when I edit a perl library used by the CGI?
           Do a touch on the main cgi file that is executed.  The mtime on the main file is
           checked each time the front-end runs.

       Do I need to be root to install and/or run SpeedyCGI?
           No, root is not required.

       How can I determine if my perl app needs to be changed to work with speedy?  Or is there
       no modification necessary?
           You may have to make modifications.

           Globals retain their values between runs, which can be good for keeping persistent
           database handles for example, or bad if your code assumes they're undefined.

           Also, if you create global variables with "my", you shouldn't try to reference those
           variables from within a subroutine - you should pass them into the subroutine instead.
           Or better yet just declare global variables with "use vars" instead of "my" to avoid
           the problem altogether.

           Here's a good explanation of the problem - it's for mod_perl, but the same thing
           applies to speedycgi:


           If all else fails you can disable persistence by setting MaxRuns to 1.  The only
           benefit of this over normal perl is that speedy will pre-compile your script.

       How do I keep a persistent connection to a database?
           Since globals retain their values between runs, the best way to do this is to store
           the connection in a global variable, then check on each run to see if that variable is
           already defined.

           For example, if your code has an "open_db_connection" subroutine that returns a
           database connection handle, you can use the code below to keep a persistent

               use vars qw($dbh);
               unless (defined($dbh)) {
                   $dbh = &open_db_connection;

           This code will store a persistent database connection handle in the global variable
           "$dbh" and only initialize it the first time the code is run.  During subsequent runs,
           the existing connection is re-used.

           You may also want to check the connection each time before using it, in case it is not
           working for some reason.  So, assuming you have a subroutine named "db_connection_ok"
           that returns true if the db connection is working, you can use code like this:

               use vars qw($dbh);
               unless (defined($dbh) && &db_connection_ok($dbh)) {
                   $dbh = &open_db_connection;

       Why do scripts with persistent Oracle database connections hang?
           When using an IPC connection to Oracle, an oracle process is fork'ed and exec'ed and
           keeps the stdout connection open, so that the web server never gets an EOF.  To fix
           the problem, either switch to using a TCP connection to the database, or add the
           following perl code somewhere before the DBI->connect statement:

               use Fcntl;
               fcntl(STDOUT, F_SETFD, FD_CLOEXEC);

           This will set the close-on-exec flag on standard out so it is closed when oracle is


       The group feature in SpeedyCGI can be used to help reduce the amount of memory used by the
       perl interpreters.  By default groups are not used (group name is "none"), and each perl
       script is given its own set of perl interpreters.  Each perl interpreter is also a
       separate system process.

       When grouping is used, perl interpreters and perl scripts are put in a group.  All perl
       interpreters in a group can run perl scripts in the same group.  What this means is that
       by putting all your scripts into one group, there could be one perl interpreter running
       all the perl scripts on your system.  This can greatly reduce your memory needs when
       running lots of different perl scripts.

       SpeedyCGI group names are entities unto themselves.  They are not associated with Unix
       groups, or with the Group directive in Apache.  Expect for the two special group names
       "none" and "default", all group names are created by the user of SpeedyCGI using the Group
       option described in "OPTIONS"

       If you want the maximum amount of grouping possible then you should run all scripts with
       the group option set to "default".  This the group name used if you just specify "-g" on
       the command line without an explicit group name.  When you do this, you will get the
       fewest number of perl interpreters possible - any perl interpreter will be able to run any
       of your perl scripts.

       Although using group "default" for all scripts results in the most efficient use of
       resources, it's not always possible or desirable to do this.  You may want to use other
       group names for the following reasons:

       ·   To isolate misbehaving scripts, or scripts that don't work in groups.

           Some scripts won't work in groups.  When perl scripts are grouped together they are
           each given their own unique package name - they are not run out of the "main" package
           as they normally would be.  So, for example, a script that explicitly uses "main"
           somewhere in its code to find its subroutines or variables probably won't work in
           groups.  In this case, it's probably best to run such a script with group "none", so
           it's compiled and run out of package main, and always given its own interpreter.

           In other cases, scripts may make changes to included packages, etc, that may break
           other scripts running in the same interpreter.  In this case such scripts can be given
           their own group name (like "pariah") to keep them away from scripts they are
           incompatible with.  The rest of your scripts can then run out of group "default".
           This will ensure that the "pariah" scripts won't run within the same interpreter as
           the other scripts.

       ·   To pass different perl or SpeedyCGI parameters to different scripts.

           You may want to use separate groups to create different policies for different

           For example, you may have an email application that contains ten perl scripts, and
           since the common perl code used in this application has a bad memory leak, you want to
           use a MaxRuns setting of 5 for all of these scripts.  You want to run all your other
           scripts with a normal MaxRuns setting.  To accomplish this you can edit the ten email
           application scripts, and at the top use the line:

               #!/usr/bin/speedy -- -gmail -r5

           In the rest of your perl scripts you can use:

               #!/usr/bin/speedy -- -g

           What this will do is put the ten email scripts into a group of their own (named
           "mail") and give them all the default MaxRuns value of 5.  All other scripts will be
           put into the group named "default", and this group will have a normal MaxRuns setting.


       Binaries for many OSes can be found at:


   Source Code
       The standard source code distribution can be retrieved from any CPAN mirror or from:


       The latest development code can be obtained from the SourceForge CVS repository using the
       following commands:

        cvs login
        cvs -z3 co 2.x

       Press Enter when prompted for a password.


           Sam Horrocks

       A lot of people have helped out with code, patches, ideas, resources, etc.  I'm sure I'm
       missing someone here - if so, please drop me an email.

       ·   Gunther Birznieks

       ·   Diana Eichert

       ·   Takanori Kawai

       ·   Robert Klep

       ·   Marc Lehmann

       ·   James McGregor

       ·   Josh Rabinowitz

       ·   Dave Parker

       ·   Craig Sanders

       ·   Joseph Wang


       perl(1), httpd(8), apxs(8).


   SpeedyCGI Home Page

   Mailing List
       ·   SpeedyCGI users mailing list -  Archives and
           subscription information are at

       ·   SpeedyCGI announcements mailing list -
           Archives and subscription information are at

   Bugs and Todo List
       Please report any bugs or requests for changes to the mailing list.

       The current bugs / todo list can be found at  Go to the Bug Tracking menu and select
       the group "bug" for bugs, or the group "rfe" for the todo list.

   Japanese Translation


   Success Stories

   Revision History

   YAPC 2001 Presentation
       I gave an Introduction to SpeedyCGI talk at YAPC 2001.  It can be found at


       Copyright (C) 2003  Sam Horrocks

       This program is free software; you can redistribute it and/or modify it under the terms of
       the GNU General Public License as published by the Free Software Foundation; either
       version 2 of the License, or (at your option) any later version.

       This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
       without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
       See the GNU General Public License for more details.

       You should have received a copy of the GNU General Public License along with this program;
       if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
       MA  02111-1307, USA.

       This product includes software developed by the Apache Software Foundation