Home

Awesome

li3_gearman offers integration between [the most RAD PHP framework] lithium and an excelent framework to farm out work to machines known as Gearman gearman.

License

li3_gearman is released under the [BSD License] license.

Installation

It is recommended that you install li3_gearman as a GIT submodule, in order to keep up with the latest upgrades. To do so, switch to the core directory holding your lithium application, and do:

$ git submodule add https://github.com/mariano/li3_gearman.git libraries/li3_gearman

Once you have downloaded li3_gearman and placed it in your main libraries folder, or your app/libraries folder, you need to enable it by placing the following at the end of your app/config/bootstrap/libraries.php file:

Libraries::add('li3_gearman');

You obviously need a Gearman server running somewhere.

Usage

There are two main elements to li3_gearman: the daemon (a lithium console command named gearmand), and the utility class used to trigger jobs.

Both tools require you to define a configuration, where you specify which Gearman servers are available. You do so by using the config() method of the Gearman class somewhere in your boostrap process. For example, to define a default configuration that uses a Gearman server located in th same server as your lithium application, add the following code to app/config/bootstrap/connections.php:

\li3_gearman\Gearman::config(array(
    'default' => array(
        'servers' => '127.0.0.1'
    )
));

Once you have a valid configuration, you can start the daemon, and start triggering jobs.

Using supervisor to run the daemon

The recommended way to run li3_gearman daemon is to use supervisor supervisor, or a similar tool. As an example, here's a supervisor configuration that runs 8 workers (assuming the application lives at /var/www)

[program:worker]
user=ubuntu
command=/usr/bin/php -f /var/www/libraries/lithium/console/lithium.php -- --app=/var/www gearmand work --blocking --verbose
process_name=%(program_name)s #%(process_num)s
numprocs=8
autostart=true
autorestart=true
stdout_logfile=/var/log/supervisor/worker-%(process_num)s.log
stdout_logfile_maxbytes=50MB
stdout_logfile_backups=10
stderr_logfile=/var/log/supervisor/worker-%(process_num)s-error.log
stderr_logfile_maxbytes=50MB
stderr_logfile_backups=10

Manually running the daemon

The daemon is a lithium console command called gearmand. Running it without arguments will show a message similar to this:

USAGE
    li3 gearmand start [<config>]
    li3 gearmand shutdown
    li3 gearmand restart
DESCRIPTION
    Gearman daemon implementation in Lithium.
OPTIONS
    ping
        Test that at least a worker is working
    start
        Start the daemon using the given configuration.
    shutdown
        Stop the daemon. Only applicable if started in daemon mode.
    restart
        Restart the daemon. Only applicable if started in daemon mode.
    --environment=<string>
        Override environment to work on. Defaults to: environment set in bootstrap
    --atomic
        If enabled, once a worker has performed its job, it will quit and a new
        worker will be spawned (thus this setting means that `resuscitate` is
        automatically enabled, and `limit` set to 0). Default: enabled
    --blocking
        Enable to interact with Gearman in blocking mode. Default: disabled
    --daemon
        Enable to start daemon in a new process. Default: disabled
    --limit=<int>
        How many workers (in total) are allowed to be spawned before finishing
        daemon. Set to 0 to not limit spawned worker count. Default: 8
    --pid=<string>
        Location of PID file. Only applicable if daemon mode is enabled.
        Default: /var/run/li3_gearman.pid
    --resuscitate
        If enabled, there will always be the number of workers defined in the
        setting "workers". If a worker dies, another one will take its place,
        up until the "limit" setting is reached. If disabled, no new
        workers will be spawned after the initial set is started.
        Default: disabled
    --verbose
        Enable to print out debug messages. If not enabled, messages go to
        user's syslog (usually /var/log/user.log). Default: disabled
    --workers=<int>
        How many workers to run. Default: 4

The daemon can be run in full daemon mode (meaning that a new process, detached from the console, is created to handle the workers), or in interactive mode. By default it will run in interactive mode, but you can switch to deaemon mode using the --daemon option.

Whatever method you decide to use, note that a certain amount of workers (processes) are spawned. This number is governed by the option --workers. If you wish to ensure that there is always a certain number of workers active and handling jobs, use the option --resuscitate, which will periodically loop through the pool to respawn workers whenever other workers are finished, for any reason. You can also limit the total number of workers which are spawned as a result of resuscitation with the option --limit.

Testing

Once you have launched the daemon, you can quickly test that the setup is working properly via the ping command. For example:

$ li3 gearmand ping --verbose
Pinging... OK

Adding an Initscript

If you want to have li3_gearman start at boottime and shutdown accordingly, you may want to look into init scripts for your local system. If you are using [Initscripts] linux-initscripts, it gets pretty easy. Create a file named li3_gearmand, make it executable (chmod +x li3_gearmand) and place it in your initscripts directory (usually either /etc/rc.d or /etc/init.d) with the following contents:

Adding the initscript, and running the necessary commands so your initscript is auto-started is definitiely not the only thing you should do. You should make sure you have some sort of monitoring tool so if the daemon falls to pieces, or no active workers are waiting for jobs, you can restart the daemon. A great tool for this is monit monit, check it out.

#!/bin/bash

# Uncomment and change the line below to match the app/ directory of your
# lithium application

#li3_app=/path/to/app

. /etc/rc.conf
. /etc/rc.d/functions

if [ -z "$li3_app" ]; then
    echo "You must set \$li3_app to point to your lithium's app/ folder"
    stat_fail
    exit 1
fi

if [ -z "$li3" ]; then
    li3_core=${li3_app%/*}
    if [ -x "li3" ]; then
        li3="li3"
    elif [ -x "$li3_app/app/libraries/lithium/console/li3" ]; then
        li3="$li3_app/app/libraries/lithium/console/li3"
    elif [ -x "$li3_core/libraries/lithium/console/li3" ]; then
        li3="$li3_core/libraries/lithium/console/li3"
    fi
fi

if [ -z "$li3" -o ! -x "$li3" ]; then
    echo $li3
    echo "You must set \$li3 to point to the li3 console binary"
    stat_fail
    exit 1
fi

daemon_bin="$li3 --app=$li3_app gearmand"
daemon_name=$(basename $0)
PIDF="/var/run/$daemon_name.pid"

get_pid() {
    [ -f $PIDF ] && cat $PIDF
}

case "$1" in
  start)
    stat_busy "Starting $daemon_name daemon"

    PID=$(get_pid)
    if [ -z "$PID" ]; then
      [ -f $PIDF ] && rm -f $PIDF
      $daemon_bin start --blocking --daemon --pid=$PIDF
      if [ $? -gt 0 ]; then
        stat_fail
        exit 1
      else
        add_daemon $daemon_name
        stat_done
      fi
    else
      stat_fail
      exit 1
    fi
    ;;

  stop)
    stat_busy "Stopping $daemon_name daemon"
    PID=$(get_pid)
    # KILL
    [ ! -z "$PID" ] && kill $PID &> /dev/null
    #
    if [ $? -gt 0 ]; then
      stat_fail
      exit 1
    else
      rm_daemon $daemon_name
      rm -f $PIDF &> /dev/null
      stat_done
    fi
    ;;

  restart)
    $0 stop
    sleep 3
    $0 start
    ;;

  status)
    stat_busy "Checking $daemon_name status";
    ck_status $daemon_name
    ;;

  *)
    echo "usage: $0 {start|stop|restart|status}"
esac

exit 0

Make sure you change the variable $li3_app to match your application's app/ directory. Once you've done that, you can start / restart / shutdown the daemon the usual initscripts way. For example, you can start the daemon with:

$ rc.d start li3_gearmand

Triggering jobs

Jobs can be triggered using the Gearman::run() method. This method takes the following arguments:

When jobs are triggered, one of the available workers (spawned by li3_gearman's gearmand daemon) will handle and execute it from within the console. This means that any resource available as part of lithium's dispatch cycle is immediately available to the job. So you can use your models, plugins, everything!

Example

Before starting with this example, make sure you have installed li3_gearman as clarified in the section Installation, and that you have added a default configuration as exemplified at the beginning of the section Usage.

Let us start by creating a task called Hello. Create a folder named tasks and place it in your application's app folder. Inside that folder, create a file named Hello.php with the following contents:

<?php
namespace app\tasks;

class Hello {
    public static function run() {
        echo "I am " . __METHOD__ . "\n";
    }

    public static function say($name) {
        echo "I am " . __METHOD__ . "\n";
        return 'Hello ' . $name;
    }
}
?>

Now, let us add the code to trigger some tasks, both in background and synchronous mode. Create a file named TestHelloController.php and place it in your app/controllers folder with the following contents:

<?php
namespace app\controllers;

use \li3_gearman\Gearman;

class TestHelloController extends \lithium\action\Controller {
    public function index() {
        $result = Gearman::run('default', 'app\tasks\Hello');
        var_dump($result);

        $result = Gearman::run('default', 'app\tasks\Hello::say', array(
            'Mariano'
        ), array('background' => false));
        var_dump($result);

        $this->_stop();
    }
}
?>

Ok we are now ready to start the daemon. For the purpose of testing (at the time of this writing, the PHP Gearman extension has [some issues] php-bug-60764 when running in non blocking mode) I'll run this in blocking mode. Since we also want some output, we'll ask the daemon to be verbose. Standing in your lithium core directory, run:

$ li3 gearmand start --verbose --blocking

You should see an output similar to the following (and the console should not return to the prompt, since we are not running in daemon mode):

li3_gearman[6331]: Daemon started with PID 6331
li3_gearman[6331]: (Daemon) Created worker number 1 with PID 6332
li3_gearman[6332]: (Worker) Starting worker
li3_gearman[6332]: (Worker) Creating Gearman worker
li3_gearman[6332]: (Worker) Registering function li3_gearman\extensions\command\Gearmand::run
li3_gearman[6331]: (Daemon) Created worker number 2 with PID 6333
li3_gearman[6333]: (Worker) Starting worker
li3_gearman[6333]: (Worker) Creating Gearman worker
li3_gearman[6333]: (Worker) Registering function li3_gearman\extensions\command\Gearmand::run
li3_gearman[6331]: (Daemon) Created worker number 3 with PID 6334
li3_gearman[6334]: (Worker) Starting worker
li3_gearman[6334]: (Worker) Creating Gearman worker
li3_gearman[6334]: (Worker) Registering function li3_gearman\extensions\command\Gearmand::run
li3_gearman[6331]: (Daemon) Created worker number 4 with PID 6335
li3_gearman[6335]: (Worker) Starting worker
li3_gearman[6335]: (Worker) Creating Gearman worker
li3_gearman[6335]: (Worker) Registering function li3_gearman\extensions\command\Gearmand::run

Now open up a browser, and access the controller we just created. If your application lives in localhost, then browse to http://localhost/test_hello. You should see the following output on the browser:

string 'H:eternauta:27' (length=14)
string 'Hello Mariano' (length=13)

The first value shown there corresponds to the first job trigger, which ran in background mode. This means that the given value is actually a job handle. The second value corresponds to the job that was run in synchronous mode, so this is actually a value returned by the job itself!

If you switch to the console, you should see the following output as a result of the controller action we just ran:

li3_gearman[6473]: (Worker) Handling job
li3_gearman[6471]: (Worker) Handling job
I am app\tasks\Hello::run
I am app\tasks\Hello::say

Scheduled / Delayed Gearman jobs

Gearman does not offer support for out-of-the-box delayed scheduling of jobs. Every job you send to Gearman is considered to be available for immediate execution. The common consensus is that delayed tasks are application-specific logic, so li3_gearmand offers them.

To do so, Redis is used as the backend to hold scheduled jobs. When you configure the Gearman instance (see Usage section above), make sure you specify how to reach your Redis server, like so:

\li3_gearman\Gearman::config(array(
    'default' => array(
        'servers' => '127.0.0.1',
        'redis' => array(
            'host' => '127.0.0.1',
            'port' => 6379
        )
    )
));

If you now open a shell and issue:

$ li3 gearmand scheduler --verbose

You will see a process waiting for scheduled tasks. From your application code you can now set the schedule option parameter to be any DateTime (UTC) on which you want a Gearman task to be executed. Following our example above, say we want to run the Hello task one hour from now:

$result = Gearman::run('default', 'app\tasks\Hello', array(), array( 'schedule' => new \DateTime('now +1 hour', new \DateTimeZone('UTC')) ));

Approximately one hour later, the scheduler process would output something like the following:

$ li3 gearmand scheduler --verbose
Waiting for scheduled tasks
Job #1cb52faa344dc68bb178a15aff388cee04d9bf51 moved for immediate execution

And you will see this job being executed in the main Gearmand work console.

Once you are ready to deploy the scheduler process, ensure you do so with supervisor supervisor, or any other tool that keeps the process alive.