=head1 NAME
Server Life Cycle Handlers
=head1 Description
This chapter discusses server life cycle and the mod_perl handlers
participating in it.
=head1 Server Life Cycle
The following diagram depicts the Apache 2.0 server life cycle and
highlights which handlers are available to mod_perl 2.0:
=for html
Apache 2.0 starts by parsing the configuration file. After the
configuration file is parsed, the C handlers are
executed if any. After that it's a turn of C
handlers to be run. When the I phase is finished the
server immediately restarts, to make sure that it can survive graceful
restarts after starting to serve the clients.
When the restart is completed, Apache 2.0 spawns the workers that will
do the actual work. Depending on the used MPM, these can be threads,
processes or a mixture of both. For example the I MPM spawns a
number of processes, each running a number of threads. When each child
process is started C handlers are
executed. Notice that they are run for each starting process, not a
thread.
From that moment on each working thread processes connections until
it's killed by the server or the server is shutdown.
=head2 Startup Phases Demonstration Module
Let's look at the following example that demonstrates all the startup
phases:
#file:MyApache2/StartupLog.pm
#----------------------------
package MyApache2::StartupLog;
use strict;
use warnings;
use Apache2::Log ();
use Apache2::ServerUtil ();
use Fcntl qw(:flock);
use File::Spec::Functions;
use Apache2::Const -compile => 'OK';
my $log_path = catfile Apache2::ServerUtil::server_root,
"logs", "startup_log";
my $log_fh;
sub open_logs {
my ($conf_pool, $log_pool, $temp_pool, $s) = @_;
$s->warn("opening the log file: $log_path");
open $log_fh, ">>$log_path" or die "can't open $log_path: $!";
my $oldfh = select($log_fh); $| = 1; select($oldfh);
say("process $$ is born to reproduce");
return Apache2::Const::OK;
}
sub post_config {
my ($conf_pool, $log_pool, $temp_pool, $s) = @_;
say("configuration is completed");
return Apache2::Const::OK;
}
sub child_init {
my ($child_pool, $s) = @_;
say("process $$ is born to serve");
return Apache2::Const::OK;
}
sub child_exit {
my ($child_pool, $s) = @_;
say("process $$ now exits");
return Apache2::Const::OK;
}
sub say {
my ($caller) = (caller(1))[3] =~ /([^:]+)$/;
if (defined $log_fh) {
flock $log_fh, LOCK_EX;
printf $log_fh "[%s] - %-11s: %s\n",
scalar(localtime), $caller, $_[0];
flock $log_fh, LOCK_UN;
}
else {
# when the log file is not open
warn __PACKAGE__ . " says: $_[0]\n";
}
}
my $parent_pid = $$;
END {
my $msg = "process $$ is shutdown";
$msg .= "\n". "-" x 20 if $$ == $parent_pid;
say($msg);
}
1;
And the I configuration section:
StartServers 4
MinSpareServers 4
MaxSpareServers 4
MaxClients 10
MaxRequestsPerChild 0
PerlModule MyApache2::StartupLog
PerlOpenLogsHandler MyApache2::StartupLog::open_logs
PerlPostConfigHandler MyApache2::StartupLog::post_config
PerlChildInitHandler MyApache2::StartupLog::child_init
PerlChildExitHandler MyApache2::StartupLog::child_exit
When we perform a server startup followed by a shutdown, the
I is created if it didn't exist already (it shares
the same directory with I and other standard log files),
and each stage appends to that file its log information. So when we
perform:
% bin/apachectl start && bin/apachectl stop
the following is getting logged to I:
[Sun Jun 6 01:50:06 2004] - open_logs : process 24189 is born to reproduce
[Sun Jun 6 01:50:06 2004] - post_config: configuration is completed
[Sun Jun 6 01:50:07 2004] - END : process 24189 is shutdown
--------------------
[Sun Jun 6 01:50:08 2004] - open_logs : process 24190 is born to reproduce
[Sun Jun 6 01:50:08 2004] - post_config: configuration is completed
[Sun Jun 6 01:50:09 2004] - child_init : process 24192 is born to serve
[Sun Jun 6 01:50:09 2004] - child_init : process 24193 is born to serve
[Sun Jun 6 01:50:09 2004] - child_init : process 24194 is born to serve
[Sun Jun 6 01:50:09 2004] - child_init : process 24195 is born to serve
[Sun Jun 6 01:50:10 2004] - child_exit : process 24193 now exits
[Sun Jun 6 01:50:10 2004] - END : process 24193 is shutdown
[Sun Jun 6 01:50:10 2004] - child_exit : process 24194 now exits
[Sun Jun 6 01:50:10 2004] - END : process 24194 is shutdown
[Sun Jun 6 01:50:10 2004] - child_exit : process 24195 now exits
[Sun Jun 6 01:50:10 2004] - child_exit : process 24192 now exits
[Sun Jun 6 01:50:10 2004] - END : process 24192 is shutdown
[Sun Jun 6 01:50:10 2004] - END : process 24195 is shutdown
[Sun Jun 6 01:50:10 2004] - END : process 24190 is shutdown
--------------------
First of all, we can clearly see that Apache always restart itself
after the first I phase is over. The logs show that the
I phase is preceded by the I phase. Only after
Apache has restarted itself and has completed the I and
I phase again, the I phase is run for each
child process. In our example we have had the setting
C, therefore you can see four child processes were
started.
Finally you can see that on server shutdown, the I phase
is run for each child process and the C block is executed by
the parent process and each of the child processes. This is because
that C block was inherited from the parent on fork.
However the presented behavior varies from MPM to MPM. This
demonstration was performed using prefork mpm. Other MPMs like winnt,
may run I and I more than once. Also the END
blocks may be run more times, when threads are involved. You should be
very careful when designing features relying on the phases covered in
this chapter if you plan support multiple MPMs. The only thing that's
sure is that you will have each of these phases run at least once.
Apache also specifies the I phase, which is executed
before the configuration files are parsed, but this is of no use to
mod_perl, because mod_perl is loaded only during the configuration
phase.
Now let's discuss each of the mentioned startup handlers and their
implementation in the C module in detail.
=head2 C
The I phase happens just before the I phase.
Handlers registered by C are usually used for
opening module-specific log files (e.g., httpd core and mod_ssl open
their log files during this phase).
At this stage the C stream is not yet redirected to
I, and therefore any messages to that stream will be
printed to the console the server is starting from (if such exists).
This phase is of type
C>.
The handler's configuration scope is
C>.
B
The I handler is passed four arguments: the configuration pool,
the logging stream pool, the temporary pool and the main server object.
The pool arguments are:
=over
=item *
C<$conf_pool> is the main process sub-pool, therefore its life-span
is the same as the main process's one. The main process is a sub-pool
of the global pool.
=item *
C<$log_pool> is a global pool's sub-pool, therefore its life-span is
the same as the Apache program's one.
META: what is it good for if it lives the same life as conf pool?
=item *
C<$temp_pool> is a C<$conf_pool> subpool, created before the config
phase, lives through the open_logs phase and get destroyed after the
post_config phase. So you will want to use that pool for doing
anything that can be discarded before the requests processing starts.
=back
All three pool arguments are instances of C>.
C<$s> is the base server object (an instance of C>).
B
The handler should return C>
if it completes successfully.
B
sub open_logs {
my ($conf_pool, $log_pool, $temp_pool, $s) = @_;
$s->warn("opening the log file: $log_path");
open $log_fh, ">>$log_path" or die "can't open $log_path: $!";
my $oldfh = select($log_fh); $| = 1; select($oldfh);
say("process $$ is born to reproduce");
return Apache2::Const::OK;
}
In our example the handler opens a log file for appending and sets the
filehandle to unbuffered mode. It then logs the fact that it's running
in the parent process.
As you've seen in the example this handler is configured by adding to
the top level of I:
PerlOpenLogsHandler MyApache2::StartupLog::open_logs
This handler can be executed only by the main server. If you want to
traverse the configured virtual hosts, you can accomplish that using a
simple loop. For example to print out the configured port numbers do:
use Apache2::ServerRec ();
# ...
sub open_logs {
my ($conf_pool, $log_pool, $temp_pool, $s) = @_;
my $port = $s->port;
warn "base port: $port\n";
for (my $vs = $s->next; $vs; $vs = $vs->next) {
my $port = $vs->port;
warn "vhost port: $port\n";
}
return Apache2::Const::OK;
}
=head2 C
The I phase happens right after Apache has processed the
configuration files, before any child processes were spawned (which
happens at the I phase).
This phase can be used for initializing things to be shared between
all child processes. You can do the same in the startup file, but in
the I phase you have an access to a complete
configuration tree (via
C>).
This phase is of type
C>.
The handler's configuration scope is
C>.
B
Arguments are exactly as for C>.
B
If the handler completes successfully it should return C>.
B
In our C>
example we used the I handler:
sub post_config {
my ($conf_pool, $log_pool, $temp_pool, $s) = @_;
say("configuration is completed");
return Apache2::Const::OK;
}
As you can see, its arguments are identical to the
I> phase's handler. In this
example handler we don't do much, but logging that the configuration
was completed and returning right away.
As you've seen in the example this handler is configured by adding to
I:
PerlPostConfigHandler MyApache2::StartupLog::post_config
Everything that applies to
C> identically applies
to this handler.
The
C>
includes another useful example.
=head2 C
The I phase happens immediately after the child process is
spawned. Each child process (not a thread!) will run the hooks of this
phase only once in their life-time.
In the prefork MPM this phase is useful for initializing any data
structures which should be private to each process. For example
C pre-opens database connections during this phase and
C sets the process' resources limits.
This phase is of type
C>.
The handler's configuration scope is
C>.
B
The I handler is passed two arguments: the child process
pool (C>) and the server object (C>).
B
If the handler completes successfully it should return C>.
B
In our C>
example we used the I handler:
sub child_init {
my ($child_pool, $s) = @_;
say("process $$ is born to serve");
return Apache2::Const::OK;
}
The example handler logs the pid of the child process it's run in and returns.
As you've seen in the example this handler is configured by adding to
I:
PerlChildInitHandler MyApache2::StartupLog::child_init
=head2 C
Opposite to the I phase, the I phase is
executed before the child process exits. Notice that it happens only
when the process exits, not the thread (assuming that you are using a
threaded mpm).
This phase is of type
C>.
The handler's configuration scope is
C>.
B
The I handler accepts two arguments: the child process
pool (C>) and the server object (C>).
B
If the handler completes successfully it should return C>.
B
In our C>
example we used the I handler:
sub child_exit {
my ($child_pool, $s) = @_;
say("process $$ now exits");
return Apache2::Const::OK;
}
The example handler logs the pid of the
child process it's run in and returns.
As you've seen in the example this handler is configured by adding to
I:
PerlChildExitHandler MyApache2::StartupLog::child_exit
=head1 Apache Command-line Commands
Some notes on how Apache start/restart Apache commands affect
mod_perl.
META: not sure this is the best place for this section, but start some
notes here.
Apache re-parses F at least once for B of the
following commands (and will run any mod_perl code found in it).
=over
=item httpd -k start
No special issues here.
Apache start and immediately restarts itself.
=item httpd -k restart
This will abort any processed requests and restart the server.
All kind of problems could be encountered here, including segfaults
and other kind of crashes. This is because when the C signal
is sent, things in process will be aborted.
Avoid using this method.
Alternatively C can be executed C.
=item httpd -k graceful
No issues here. Apache starts and restarts itself just like with
C, but it waits for the existing requests to finish before
killing them.
Alternatively C can be executed C.
=item httpd -k stop
Similarly to C you may encounter all kind of issues
here, due to the C signal.
=back
=head1 mod_perl Startup
The following sections discuss the specifics of the mod_perl startup.
=head2 Start Immediately Restarts
As explained in the L, on start Apache normally
runs the server configuration phase, followed by
C>
and
C>
phases, then immediately restarts itself. Therefore everything running
at the server startup is executed twice. During the restart, Perl is
completely destroyed and started again.
=head2 When Does perl Start To Run
If Apache is started as C<'httpd -t'> (equivalent to C<'apachectl
configtest'>) or as C<'httpd -S'>, it will run only the configuration
phase and exit. Depending on your configuration file, it may or may
not start perl. See the details below.
During the normal startup, mod_perl 2.0 postpones the startup of perl
until after the configuration phase is over, to allow the usage of the
C>
directive, which can't be used after Perl is started.
After the configuration phase is over, as the very first thing during
the C>,
mod_perl starts perl and runs any registered
C> and
C>
entries.
At the very end of the C> any
registrered
C>
entries are run.
When any of the following configuration directives is encountered
(during the configuration phase) mod_perl 2.0 is forced to start as
soon as they are encountered (as these options require a running
perl):
=over
=item *
C>
=item *
CPerlE section|docs::2.0::api::Apache2::PerlSections>>
=item *
C>
=back
Therefore if you want to trigger an early Perl startup, you could add
an empty CPerlE
section|docs::2.0::api::Apache2::PerlSections>> in F:
# trigger an early Perl startup
right after loading the mod_perl module, if you are using DSO, or just
before your mod_perl configuration section, if you're using a static
mod_perl build. But most likely you want to use the
C>
instead.
=head2 Startup File
A startup file with Perl code to be executed at the server startup can
be loaded using
C>. For
example:
PerlPostConfigRequire /home/httpd/perl/lib/startup.pl
It's used to adjust Perl modules search paths in C<@INC>, pre-load
commonly used modules, pre-compile constants, etc. Here is a typical
I for mod_perl 2.0:
#file:startup.pl
#---------------
use lib qw(/home/httpd/perl);
# enable if the mod_perl 1.0 compatibility is needed
# use Apache2::compat ();
# preload all mp2 modules
# use ModPerl::MethodLookup;
# ModPerl::MethodLookup::preload_all_modules();
use ModPerl::Util (); #for CORE::GLOBAL::exit
use Apache2::RequestRec ();
use Apache2::RequestIO ();
use Apache2::RequestUtil ();
use Apache2::ServerRec ();
use Apache2::ServerUtil ();
use Apache2::Connection ();
use Apache2::Log ();
use APR::Table ();
use ModPerl::Registry ();
use Apache2::Const -compile => ':common';
use APR::Const -compile => ':common';
1;
In this file C<@INC> in adjusted to include non-standard
directories with Perl modules:
use lib qw(/home/httpd/perl);
If you need to use the backwards compatibility layer load:
use Apache2::compat ();
Next we preload the commonly used mod_perl 2.0 modules and precompile
common constants.
Finally as usual the I file must be terminated with C<1;>.
=head2 Dealing with Restarts
Ideally the code running at the server startup shouldn't be affected
by L. If however this
is not the case, you can use
C>.
=head1 Maintainers
Maintainer is the person(s) you should contact with updates,
corrections and patches.
=over
=item *
Stas Bekman [http://stason.org/]
=back
=head1 Authors
=over
=item *
=back
Only the major authors are listed above. For contributors see the
Changes file.
=cut