More cooked

[catagits/Catalyst-Manual.git] / lib / Catalyst / Manual / Deployment.pod

=head1 NAME

Catalyst::Manual::Deployment - Deploying Catalyst

=head1

=head1 mod_perl

L<Catalyst::Manual::Deployment::Apache::mod_perl>

=head1 FastCGI

=head2 Apache

L<Catalyst::Manual::Deployment::Apache::FastCGI>


=head2 mod_perl Deployment

mod_perl is not the best solution for many applications, but we'll list some
pros and cons so you can decide for yourself. The other (recommended)
deployment option is FastCGI, for which see below.

=head3 Pros

=head4 Speed

mod_perl is fast and your app will be loaded in memory
within each Apache process.

=head4 Shared memory for multiple apps

If you need to run several Catalyst apps on the same server, mod_perl will
share the memory for common modules.

=head3 Cons

=head4 Memory usage

Since your application is fully loaded in memory, every Apache process will
be rather large.  This means a large Apache process will be tied up while
serving static files, large files, or dealing with slow clients.  For this
reason, it is best to run a two-tiered web architecture with a lightweight
frontend server passing dynamic requests to a large backend mod_perl
server.

=head4 Reloading

Any changes made to the core code of your app require a full Apache restart.
Catalyst does not support Apache::Reload or StatINC.  This is another good
reason to run a frontend web server where you can set up an
C<ErrorDocument 502> page to report that your app is down for maintenance.

=head4 Cannot run multiple versions of the same app

It is not possible to run two different versions of the same application in
the same Apache instance because the namespaces will collide.

=head4 Cannot run different versions of libraries.

If you have two different applications which run on the same machine,
which need two different versions of a library then the only way to do
this is to have per-vhost perl interpreters (with different library paths).
This is entirely possible, but nullifies all the memory sharing benefits that
you get from having multiple applications sharing the same interpreter.

=head4 Setup

Now that we have that out of the way, let's talk about setting up mod_perl
to run a Catalyst app.

=head4 1. Install Catalyst::Engine::Apache

You should install the latest versions of both Catalyst and
Catalyst::Engine::Apache.  The Apache engines were separated from the
Catalyst core in version 5.50 to allow for updates to the engine without
requiring a new Catalyst release.

=head4 2. Install Apache with mod_perl

Both Apache 1.3 and Apache 2 are supported, although Apache 2 is highly
recommended.  With Apache 2, make sure you are using the prefork MPM and not
the worker MPM.  The reason for this is that many Perl modules are not
thread-safe and may have problems running within the threaded worker
environment.  Catalyst is thread-safe however, so if you know what you're
doing, you may be able to run using worker.

In Debian, the following commands should get you going.

    apt-get install apache2-mpm-prefork
    apt-get install libapache2-mod-perl2

=head4 3. Configure your application

Every Catalyst application will automagically become a mod_perl handler
when run within mod_perl.  This makes the configuration extremely easy.
Here is a basic Apache 2 configuration.

    PerlSwitches -I/var/www/MyApp/lib
    PerlModule MyApp

    <Location />
        SetHandler          modperl
        PerlResponseHandler MyApp
    </Location>

The most important line here is C<PerlModule MyApp>.  This causes mod_perl
to preload your entire application into shared memory, including all of your
controller, model, and view classes and configuration.  If you have -Debug
mode enabled, you will see the startup output scroll by when you first
start Apache.

For an example Apache 1.3 configuration, please see the documentation for
L<Catalyst::Engine::Apache::MP13>.

=head3 Test It

That's it, your app is now a full-fledged mod_perl application!  Try it out
by going to http://your.server.com/.

=head3 Other Options

=head4 Non-root location

You may not always want to run your app at the root of your server or virtual
host.  In this case, it's a simple change to run at any non-root location
of your choice.

    <Location /myapp>
        SetHandler          modperl
        PerlResponseHandler MyApp
    </Location>

When running this way, it is best to make use of the C<uri_for> method in
Catalyst for constructing correct links.

=head4 Static file handling

Static files can be served directly by Apache for a performance boost.

    DocumentRoot /var/www/MyApp/root
    <Location /static>
        SetHandler default-handler
    </Location>

This will let all files within root/static be handled directly by Apache.  In
a two-tiered setup, the frontend server should handle static files.
The configuration to do this on the frontend will vary.

The same is accomplished in lighttpd with the following snippet:

   $HTTP["url"] !~ "^/(?:img/|static/|css/|favicon.ico$)" {
         fastcgi.server = (
             "" => (
                 "MyApp" => (
                     "socket"       => "/tmp/myapp.socket",
                     "check-local"  => "disable",
                 )
             )
         )
    }

Which serves everything in the img, static, css directories
statically, as well as the favicon file.

Note the path of the application needs to be stated explicitly in the
web server configuration for both these recipes.

=head2 Catalyst on shared hosting

So, you want to put your Catalyst app out there for the whole world to
see, but you don't want to break the bank. There is an answer - if you
can get shared hosting with FastCGI and a shell, you can install your
Catalyst app in a local directory on your shared host. First, run

    perl -MCPAN -e shell

and go through the standard CPAN configuration process. Then exit out
without installing anything. Next, open your .bashrc and add

    export PATH=$HOME/local/bin:$HOME/local/script:$PATH
    perlversion=`perl -v | grep 'built for' | awk '{print $4}' | sed -e 's/v//;'`
    export PERL5LIB=$HOME/local/share/perl/$perlversion:$HOME/local/lib/perl/$perlversion:$HOME/local/lib:$PERL5LIB

and log out, then back in again (or run C<". .bashrc"> if you
prefer). Finally, edit C<.cpan/CPAN/MyConfig.pm> and add

    'make_install_arg' => qq[SITEPREFIX=$ENV{HOME}/local],
    'makepl_arg' => qq[INSTALLDIRS=site install_base=$ENV{HOME}/local],

Now you can install the modules you need using CPAN as normal; they
will be installed into your local directory, and perl will pick them
up. Finally, change directory into the root of your virtual host and
symlink your application's script directory in:

    cd path/to/mydomain.com
    ln -s ~/lib/MyApp/script script

And add the following lines to your .htaccess file (assuming the server
is setup to handle .pl as fcgi - you may need to rename the script to
myapp_fastcgi.fcgi and/or use a SetHandler directive):

  RewriteEngine On
  RewriteCond %{REQUEST_URI} !^/?script/myapp_fastcgi.pl
  RewriteRule ^(.*)$ script/myapp_fastcgi.pl/$1 [PT,L]

Now C<http://mydomain.com/> should now Just Work. Congratulations, now
you can tell your friends about your new website (or in our case, tell
the client it's time to pay the invoice :) )

=head2 FastCGI Deployment

FastCGI is a high-performance extension to CGI. It is suitable
for production environments.

=head3 Pros

=head4 Speed

FastCGI performs equally as well as mod_perl.  Don't let the 'CGI' fool you;
your app runs as multiple persistent processes ready to receive connections
from the web server.

=head4 App Server

When using external FastCGI servers, your application runs as a standalone
application server.  It may be restarted independently from the web server.
This allows for a more robust environment and faster reload times when
pushing new app changes.  The frontend server can even be configured to
display a friendly "down for maintenance" page while the application is
restarting.

=head4 Load-balancing

You can launch your application on multiple backend servers and allow the
frontend web server to load-balance between all of them.  And of course, if
one goes down, your app continues to run fine.

=head4 Multiple versions of the same app

Each FastCGI application is a separate process, so you can run different
versions of the same app on a single server.

=head4 Can run with threaded Apache

Since your app is not running inside of Apache, the faster mpm_worker module
can be used without worrying about the thread safety of your application.

=head3 Cons

You may have to disable mod_deflate.  If you experience page hangs with
mod_fastcgi then remove deflate.load and deflate.conf from mods-enabled/

=head4 More complex environment

With FastCGI, there are more things to monitor and more processes running
than when using mod_perl.

=head3 Setup

=head4 1. Install Apache with mod_fastcgi

mod_fastcgi for Apache is a third party module, and can be found at
L<http://www.fastcgi.com/>.  It is also packaged in many distributions,
for example, libapache2-mod-fastcgi in Debian. You will also need to install
the L<FCGI> module from cpan.

Important Note! If you experience difficulty properly rendering pages,
try disabling Apache's mod_deflate (Deflate Module), e.g. 'a2dismod deflate'.

=head4 2. Configure your application

    # Serve static content directly
    DocumentRoot  /var/www/MyApp/root
    Alias /static /var/www/MyApp/root/static

    FastCgiServer /var/www/MyApp/script/myapp_fastcgi.pl -processes 3
    Alias /myapp/ /var/www/MyApp/script/myapp_fastcgi.pl/

    # Or, run at the root
    Alias / /var/www/MyApp/script/myapp_fastcgi.pl/

The above commands will launch 3 app processes and make the app available at
/myapp/

=head3 Standalone server mode

While not as easy as the previous method, running your app as an external
server gives you much more flexibility.

First, launch your app as a standalone server listening on a socket.

    script/myapp_fastcgi.pl -l /tmp/myapp.socket -n 5 -p /tmp/myapp.pid -d

You can also listen on a TCP port if your web server is not on the same
machine.

    script/myapp_fastcgi.pl -l :8080 -n 5 -p /tmp/myapp.pid -d

You will probably want to write an init script to handle starting/stopping
of the app using the pid file.

Now, we simply configure Apache to connect to the running server.

    # 502 is a Bad Gateway error, and will occur if the backend server is down
    # This allows us to display a friendly static page that says "down for
    # maintenance"
    Alias /_errors /var/www/MyApp/root/error-pages
    ErrorDocument 502 /_errors/502.html

    FastCgiExternalServer /tmp/myapp.fcgi -socket /tmp/myapp.socket
    Alias /myapp/ /tmp/myapp.fcgi/

    # Or, run at the root
    Alias / /tmp/myapp.fcgi/

=head3 More Info

L<Catalyst::Engine::FastCGI>.

=head2 Development server deployment

The development server is a mini web server written in perl.  If you
expect a low number of hits or you don't need mod_perl/FastCGI speed,
you could use the development server as the application server with a
lightweight proxy web server at the front.  However, consider using
L<Catalyst::Engine::HTTP::Prefork> for this kind of deployment instead, since
it can better handle multiple concurrent requests without forking, or can
prefork a set number of servers for improved performance.

=head3 Pros

As this is an application server setup, the pros are the same as
FastCGI (with the exception of speed).
It is also:

=head4 Simple

The development server is what you create your code on, so if it works
here, it should work in production!

=head3 Cons

=head4 Speed

Not as fast as mod_perl or FastCGI. Needs to fork for each request
that comes in - make sure static files are served by the web server to
save forking.

=head3 Setup

=head4 Start up the development server

   script/myapp_server.pl -p 8080 -k  -f -pidfile=/tmp/myapp.pid

You will probably want to write an init script to handle stop/starting
the app using the pid file.

=head4 Configuring Apache

Make sure mod_proxy is enabled and add:

    # Serve static content directly
    DocumentRoot /var/www/MyApp/root
    Alias /static /var/www/MyApp/root/static

    ProxyRequests Off
    <Proxy *>
        Order deny,allow
        Allow from all
    </Proxy>

    # Need to specifically stop these paths from being passed to proxy
    ProxyPass /static !
    ProxyPass /favicon.ico !

    ProxyPass / http://localhost:8080/
    ProxyPassReverse / http://localhost:8080/

    # This is optional if you'd like to show a custom error page
    # if the proxy is not available
    ErrorDocument 502 /static/error_pages/http502.html

You can wrap the above within a VirtualHost container if you want
different apps served on the same host.

=head1 AUTHORS

Catalyst Contributors, see Catalyst.pm

=head1 COPYRIGHT

This library is free software. You can redistribute it and/or modify it under
the same terms as Perl itself.

=cut