NAME
CGI::SpeedyCGI - Speed up perl CGI scripts by running them persistently
SYNOPSIS
#!/usr/bin/speedy
### Your CGI Script Here
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->set_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";
DESCRIPTION
SpeedyCGI is a way to run CGI perl scripts persistently, which usually
makes them run much more quickly. A script can be converted to SpeedyCGI
by changing the interpreter line at the top of the script from:
#!/usr/bin/perl
to
#!/usr/bin/speedy
After the script is initially run, instead of exiting, SpeedyCGI keeps
the perl interpreter running. During subsequent runs, this interpreter
is used to handle new requests instead of starting a new perl
interpreter for each execution.
SpeedyCGI conforms to the CGI specification, and normally does not work
inside the web server. A very fast cgi-bin, written in C, is executed
for each request. This fast cgi-bin then contacts the persistent Perl
process, which is usually already running, to do the work and return the
results.
Since the perl interpreter runs outside the web server, it can't cause
problems for the web server itself. By default each perl script runs in
its own Unix process, so one program 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 persistently.
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.
OPTIONS
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
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.
CGI::SpeedyCGI
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 the section on "METHODS"
below.
mod_speedycgi
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
speedy Timeout option, use the apache directive SpeedyTimeout.
Context
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:
speedy
The command-line "speedy" program, used normally with #! at the top
of your script or from a shell prompt.
mod_speedycgi
The optional Apache mod_speedycgi module.
module
During perl execution via the CGI::SpeedyCGI module's getopt/setopt
methods.
Options Available
BackendProg
Command Line : -p<string>
Default Value : "/usr/bin/speedy_backend"
Context : mod_speedycgi, speedy
Description:
Path to the speedy backend program.
BufsizGet
Command Line : -B<number>
Default Value : 8192
Context : speedy
Description:
Use <number> bytes for the buffer that
receives data from the CGI script.
BufsizPost
Command Line : -b<number>
Default Value : 1024
Context : speedy
Description:
Use <number> bytes for the buffer that sends
data to the CGI script.
Group
Command Line : -g<string>
Default Value : "none"
Context : mod_speedycgi, speedy
Description:
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.
MaxBackends
Command Line : -M<number>
Default Value : 0 (no max)
Context : mod_speedycgi, speedy
Description:
If non-zero, limits the number of speedy
backends running for this cgi script to
<number>.
MaxRuns
Command Line : -r<number>
Default Value : 500
Context : mod_speedycgi, module, speedy
Description:
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.
PerlArgs
Command Line : N/A
Default Value : ""
Context : mod_speedycgi
Description:
Command-line options to pass to the perl
interpreter.
Timeout
Command Line : -t<number>
Default Value : 3600 (one hour)
Context : mod_speedycgi, module, speedy
Description:
If no new requests have been received after
<number> seconds, exit the persistent perl
interpreter. Zero indicates no timeout.
TmpBase
Command Line : -T<string>
Default Value : "/tmp/speedy"
Context : mod_speedycgi, speedy
Description:
Use the given prefix for creating temporary
files. This must be a filename prefix, not a
directory name.
Version
Command Line : -v
Context : speedy
Description:
Print the SpeedyCGI version and exit.
METHODS
The following methods are available in the CGI::SpeedyCGI module.
new
Create a new CGI::SpeedyCGI object.
my $sp = CGI::SpeedyCGI->new;
register_cleanup($function_ref)
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.
$sp->register_cleanup(\&cleanup_func);
set_shutdown_handler($function_ref)
Register a function 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.
$sp->set_shutdown_handler(sub {$dbh->logout});
i_am_speedy
Returns a boolean telling whether this script is running under
SpeedyCGI or not. A CGI script can run under regular perl, or under
SpeedyCGI. This method allows the script to tell which environment
it is in.
$sp->i_am_speedy;
To make your script as portable as possible, you can use the
following test to make sure both the SpeedyCGI module is available
and you are running under SpeedyCGI:
if (eval {require CGI::SpeedyCGI} && CGI::SpeedyCGI->i_am_speedy) {
Do something SpeedyCGI specific here...
setopt($optname, $value)
Set one of the SpeedyCGI options given in the section on "Options
Available". Returns the option's previous value. $optname is case-
insensitive.
$sp->setopt('TIMEOUT', 300);
getopt($optname)
Return the current value of one of the SpeedyCGI options. $optname
is case-insensitive.
$sp->getopt('TIMEOUT');
INSTALLATION
To install SpeedyCGI you will need to either download a binary package
for your OS, or compile SpeedyCGI from source code. See the section on
"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:
Linux
rpm -i <filename>
Solaris
gunzip <filename>.gz
pkgadd -d <filename>
BSD pkg_add <filename>
If you are also installing the apache module you will have to configure
Apache as documented in the section on "Apache Configuration".
Source Code Installation
To compile SpeedyCGI you will need perl 5.004 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 other
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
make test
make install
This will install the speedy and speedy_backend binaries in the same
directory where perl was installed, and the SpeedyCGI.pm 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 path.
Install in a Different Directory
To install in a different directory, change the first line in the
section on "Standard Install" to:
perl Makefile.PL PREFIX=/somewhere
This will install the binaries in /somewhere/bin and the SpeedyCGI.pm
module under /somewhere/lib.
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 mod_speedycgi.so from the mod_speedycgi directory to wherever
your apache modules are stored (try /usr/lib/apache)
* 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/mod_speedycgi.so
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
</Location>
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
</Location>
BUGS / TODO
Please report any bugs or requests for changes to speedycgi@newlug.org.
The current bugs / todo list can be found at
http://www.sourceforge.net/projects/speedycgi/. Go to the Bug Tracking
menu and select the group "bug" for bugs, or the group "rfe" for the
todo list.
FREQUENTLY ASKED QUESTIONS
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 CGI 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 a CGI 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 CGI 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 the section on "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',
`@ARGV', and `%SIG' 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:
http://perl.apache.org/faq/mod_perl_cgi.html#Variables_retain_their_value_fro
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 between requests.
How do I keep a persistent connect 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 connection:
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;
}
MAILING LIST
The mailing list address is speedycgi@newlug.org. Subscribe by sending a
message to speedycgi-request@newlug.org with the word "subscribe" in the
body.
An archive of the mailing list is at
http://newlug.org/mailArchive/speedycgi/ and mirrored at
http://daemoninc.com/SpeedyCGI/mailArchive/
DOWNLOADING
Binaries
Binaries for many OSes can be found at:
http://daemoninc.com/SpeedyCGI/CGI-SpeedyCGI/binaries
There are also some older debian packages available from:
http://www.debian.org/Packages/stable/interpreters/speedy-cgi-perl.html
http://www.debian.org/Packages/unstable/web/libapache-mod-speedycgi.html
The debian packages are not based on the latest version of SpeedyCGI.
Source Code
The standard source code distribution can be retrieved from any CPAN
mirror or from:
http://daemoninc.com/SpeedyCGI/download.html
http://www.cpan.org/modules/by-authors/id/H/HO/HORROCKS/
The latest development code can be obtained from the SourceForge CVS
repository using the following commands:
cvs -d:pserver:anonymous@cvs.SpeedyCGI.sourceforge.net:/cvsroot/SpeedyCGI login
cvs -z3 -d:pserver:anonymous@cvs.SpeedyCGI.sourceforge.net:/cvsroot/SpeedyCGI co 2.x
Press Enter when prompted for a password.
AUTHOR
Sam Horrocks
Daemon Consulting Inc.
http://daemoninc.com
sam@daemoninc.com
SEE ALSO
perl(1), httpd(8), apxs(8).
TRANSLATIONS
Japanese
http://member.nifty.ne.jp/hippo2000/perltips/CGI/SpeedyCGI.htm
COPYRIGHT
Copyright (C) 2001 Daemon Consulting Inc.
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.