mod_perl

1
mod_perl

• High speed dynamic content

2
Definitions

• Apache OpenSource httpd server
• Perl OpenSource interpreted programming
  language
• mod_perl OpenSource interface between Apache
  and Perl

3
Scenario

• Unix hosted web-site using CGI Perl scripts to
  create web pages with dynamic content.
• As the site becomes larger and more popular, the
  speed slows
• The web servers are not resource constrained
• How can we speed the delivery of content?

4
Topics covered

• Why use mod_perl?
• How mod_perl works
• Installation highlights
• Commonly encountered problems when porting from
  CGI to mod_perl

5
Take home messages

• use strict
• Use the PerlTaintCheck directive
• mod_perl enabled scripts are fast
• Possibly faster than C code
• Effective use of mod_perl is not trivial

6
Why use mod_perl?

• CGI scripts written in Perl and served from
  Apache can be slow
• Generally the more dynamic the content, the
  slower the page return
• As sites grow, their content usually becomes more
  dynamic

• Sites ported to mod_perl show request return
  rates 200 to 2000 higher!!!

7
Evolution of a web-site
8
Why is CGI slow?

• For each CGI script called
• Apache must load the Perl interpreter
• Perl must load the the CGI script including all
  used Modules
• The CGI script must initialize itself on each
  load
• Database initialization is a common slow-down

9
Dynamic content options
10
The embedded interpreter

• There are increasingly sophisticated solutions
• Only load the Perl interpreter once
• (PerlRun)
• Only load the CGI scripts and modules once
• (PerlRegistry)
• Only do initializations once
• (Code re-factoring)

11
The mod_perl version

• Each Apache instance loads the Perl interpreter
  when it initializes (PerlRun)
• mod_perl loads CGI scripts the first time they
  are executed (PerlRegistry)
• We can move CGI initialization code into a
  separate script that is executed when each Apache
  instance starts (Code re-factoring)

12
How does mod_perl work?

• Its invoked by Apache so
• How does Apache work?
• Its running Perl code so
• How does Perl work?
• The mod_perl handler

13
How does Perl work?

• Interpreted language
• To run a Perl program
• The Perl interpreter is loaded
• The Program is compiled by the Perl interpreter
• The Program is executed

14
Perls _at_INC array

• Analogous to the LD_LIBRARY_PATH environment
  variable
• This array contains the paths that Perl will
  search when loading libraries or modules
• With mod_perl this is set during Apaches
  initialization phase and cannot be change later

15
Perls INC hash

• After Perl has compiled a program, it is added to
  INC as a key with the fully qualified filename
  and compile time as a values
• Before loading a program, Perl searches INC for
  a match. If it finds one, it uses the INC entry
  and does not re-load the program

16
How does Apache work?

• Apache serves documents requested via HTTP
• A sample request

17
The Apache request cycle
18
The mode _perl handlers

• You can interact with the Apache request cycle at
  any point with a handler
• A handler is a program that is defined for a
  Directory, Location, Files or FilesMatch
  directive in Apaches configuration files

19
The Apache handler API

• Apache calls the handler, passes information
  about the current transaction and server state
  and expects an integer status code in return
• If the handler returns an error, Apache
  short-circuits to the logging phase

20
The PerlHandler part 1

• Called prior to serving the requested document
• Two default handlers
• PerlRun
• PerlRegistry

21
The PerlHandler part 2

• Unless you are using a custom handler, your CGI
  programs are executed as subroutines inside the
  handler subroutine (nested subroutines)
• Since Apache never exits, the CGI programs are
  never reaped by Perls garbage collection

22
A handler subroutine

• sub handler
• my request shift(_at__)
• my status do_something(request)
• return status

23
Installation highlights

• Install Apache and mod_perl together
• Configure Apache to use mod_perl
• Port your CGI scripts

24
Installation part 1

• (The following is for Unix, Win32 installs are
  more difficult)
• Download Apache http//www.apache.org
• current version is 1.3.12
• Download mod_perl http//search.cpan.org
• current version is 1.24
• Check perl version (mod_perl 1.24 or earlier
  works best with Perl 5.00503not 5.6.0)

25
Installation part 2

• By default building mod_perl will also build
  Apache
• gtperl Makefile.PL APACHE_SRC../apache.1.3.12 \
  DO_HTTPD1 USE_APACI1 EVERYTHING1
• gtmakemake testmake install
• gtcd ../apache.1.3.12 make install

26
Configuration

• Configure Apache to use mod_perl
• modify httpd.conf

• ltIfModule mod_perl.cgt
• Include conf/perl.conf
• lt/IfModulegt

27
perl.conf part 1

• PerlSetEnv PERL5LIB /my/perl/lib
• PerlRequire conf/startup.pl
• Alias /cgi-bin /my/cgi/dir
• PerlTaintCheck On

28
perl.conf part 2

• ltLocation /cgi-bingt
• SetHandler perl-script
• PerlHandler ApacheRegistry
• Options ExecCGI
• lt/Locationgt

29
perl.conf part 3

• A good initial check of mod_perl
• ltLocation /mod_perl-statusgt
• SetHandler perl-script
• PerlHandler ApacheStatus
• lt/Locationgt

30
startup.pl

• Loads commonly used modules
• Initializes database connections
• _at_EXPORT_OK qw( dbh )
• use lib qw( /home/matt/perl/Modules
• /opt/apps/lib )
• my dbh DBI-gtconnect(db,user,passwd)

31
Porting CGI scripts to mod_perl

• use strict
• during porting
• use diagnostics And check the error.log
• Run the Apache server single threaded
• Add the following to perl.conf
• PerlInitHandler ApacheStatINC

32
Common porting problems part 1

• As someone pointed out during the presentation,
  all of these problems could be solved by using
  re-entrant coding techniques
• If you used them when you wrote the CGI scripts,
  porting will be easy
• That being said, most CGI scripts are not written
  by people familiar with such techniques

33
Common porting problems

• Run-time attempts to add paths to _at_INC
• Implicit variable passing
• Assigning anonymous sub to SIG__DIE__
• Compiled regular expressions
• No closure of file or dbm handles

34
Adding paths to _at_INC

• This will not work because _at_INC is fixed after
  the initialization of Apache
• Use startup.pl to add paths
• BEGIN blocks only get executed at compile time

35
Implicit variable passing

• The value of the variable in subroutines will not
  be updated after the first call
• Perl considers the two variables to be separate
• Solutions
• Fully qualify the variable in the subroutine
• Pass by value or reference
• Make the variable global

36
SIG__DIE__ problems

• SIG__DIE__ is global
• Your anonymous subroutine will be called by any
  die statement
• Solutions
• Write your own exception method
• Error.pm (try, throw, catch)

37
Compiled regular expressions

• Only compiled for the first pattern used
• Solution
• Add an eval block
• eval q
• foreach (_at_list)
• mysub(_) if ( /pattern/o )

38
No closure of file or dbm handles

• Common to use close on exit fh management
• Now the script does not exit.
• Solution
• Close your file and dbm handles

39
Conclusions

• mod_perl will speed up CGI scripts by 200 to
  2000
• use strict
• Use the PerlTaintCheck directive
• Effective use of mod_perl is not trivial, plan
  accordingly

40
Further Reading

• Writing Apache Modules with Perl and C
• Lincoln Stein and Doug MacEachern
• mod_perl Guide
• http//perl.apache.org/guide
• Stas Bekman
• FAQ
• http//perl.apache.org/faq
• Frank Cringle

41
Apache INC hash detail

• The handler subroutine changes the name of the
  nested subroutines to the fully qualified
  filename plus the prefix ApacheROOT
• ApacheROOTMyDirscript_2epl
• This reduces the likelihood of namespace
  collisions, but frustrates tampering