Cron is non-ideal for running scheduled application tasks, especially in an app deployed to multiple machines. More details.
Clockwork is a cron replacement. It runs as a lightweight, long-running Ruby process which sits alongside your web processes (Mongrel/Thin) and your worker processes (DJ/Resque/Minion/Stalker) to schedule recurring work at particular times or dates. For example, refreshing feeds on an hourly basis, or send reminder emails on a nightly basis, or generating invoices once a month on the 1st.
Create clock.rb:
require 'clockwork'
include Clockwork
handler do |job|
puts "Running #{job}"
end
every(10.seconds, 'frequent.job')
every(3.minutes, 'less.frequent.job')
every(1.hour, 'hourly.job')
every(1.day, 'midnight.job', :at => '00:00')
Run it with the clockwork binary:
$ clockwork clock.rb
Starting clock for 4 events: [ frequent.job less.frequent.job hourly.job midnight.job ]
Triggering frequent.job
If you would not like to taint the namespace with include Clockwork, you can use
it as the module (thanks to hoverlover).
require 'clockwork'
module Clockwork
handler do |job|
puts "Running #{job}"
end
every(10.seconds, 'frequent.job')
every(3.minutes, 'less.frequent.job')
every(1.hour, 'hourly.job')
every(1.day, 'midnight.job', :at => '00:00')
end
If you need to load your entire environment for your jobs, simply add:
require './config/boot'
require './config/environment'
under the require 'clockwork' declaration.
Clockwork fits well with heroku's cedar stack.
Consider to use clockwork-init.sh to create a new project for heroku.
The clock process only makes sense as a place to schedule work to be done, not to do the work. It avoids locking by running as a single process, but this makes it impossible to parallelize. For doing the work, you should be using a job queueing system, such as Delayed Job, Beanstalk/Stalker, RabbitMQ/Minion, Resque, or Sidekiq. This design allows a simple clock process with no locks, but also offers near infinite horizontal scalability.
For example, if you're using Beanstalk/Staker:
require 'stalker'
handler { |job| Stalker.enqueue(job) }
every(1.hour, 'feeds.refresh')
every(1.day, 'reminders.send', :at => '01:30')
Using a queueing system which doesn't require that your full application be loaded is preferable, because the clock process can keep a tiny memory footprint. If you're using DJ or Resque, however, you can go ahead and load your full application enviroment, and use per-event blocks to call DJ or Resque enqueue methods. For example, with DJ/Rails:
require 'config/boot'
require 'config/environment'
every(1.hour, 'feeds.refresh') { Feed.send_later(:refresh) }
every(1.day, 'reminders.send', :at => '01:30') { Reminder.send_later(:send_reminders) }
:at parameter the hour and minute specifies when the event occur.
The simplest example:
every(1.day, 'reminders.send', :at => '01:30')
You can omit 0 of the hour:
every(1.day, 'reminders.send', :at => '1:30')
The wildcard for hour is supported:
every(1.hour, 'reminders.send', :at => '**:30')
You can set more than one timing:
every(1.hour, 'reminders.send', :at => ['12:00', '18:00'])
# send reminders at noon and evening
You can also specify a timezone (default is the local timezone):
every(1.day, 'reminders.send', :at => '00:00', :tz => 'UTC')
# Runs the job each day at midnight, UTC.
# The value for :tz can be anything supported by [TZInfo](http://tzinfo.rubyforge.org/)
:if parameter is invoked every time the task is ready to run, and run if the
return value is true.
Run on every first day of month.
Clockwork.every(1.day, 'myjob', :if => lambda { |t| t.day == 1 })
The argument is an instance of Time. If :tz option is set, it is local time.
This argument cannot be omitted. Please use _ as placeholder if not needed.
Clockwork.every(1.second, 'myjob', :if => lambda { |_| true })
Clockwork exposes a couple of configuration options you may change:
By default Clockwork logs to STDOUT. In case you prefer to make it to use our
own logger implementation you have to specify the logger configuration option. See example below.
Clockwork wakes up once a second (by default) and performs its duties. If that
is the rare case you need to tweak the number of seconds it sleeps then you have
the sleep_timeout configuration option to set like shown below.
This is the default timezone to use for all events. When not specified this defaults to the local timezone. Specifying :tz in the the parameters for an event overrides anything set here.
Clockwork.configure do |config|
config[:sleep_timeout] = 5
config[:logger] = Logger.new(log_file_path)
config[:tz] = 'EST'
end
clock.rb is standard Ruby. Since we include the Clockwork module (the clockwork binary does this automatically, or you can do it explicitly), this exposes a small DSL ("handler" and "every") to define the handler for events, and then the events themselves.
The handler typically looks like this:
handler { |job| enqueue_your_job(job) }
This block will be invoked every time an event is triggered, with the job name passed in. In most cases, you should be able to pass the job name directly through to your queueing system.
The second part of the file are the events, which roughly resembles a crontab:
every(5.minutes, 'thing.do')
every(1.hour, 'otherthing.do')
In the first line of this example, an event will be triggered once every five minutes, passing the job name 'thing.do' into the handler. The handler shown above would thus call enqueue_your_job('thing.do').
You can also pass a custom block to the handler, for job queueing systems that rely on classes rather than job names (i.e. DJ and Resque). In this case, you need not define a general event handler, and instead provide one with each event:
every(5.minutes, 'thing.do') { Thing.send_later(:do) }
If you provide a custom handler for the block, the job name is used only for logging.
You can also use blocks to do more complex checks:
every(1.day, 'check.leap.year') do
Stalker.enqueue('leap.year.party') if Time.now.year % 4 == 0
end
Only one clock process should ever be running across your whole application deployment. For example, if your app is running on three VPS machines (two app servers and one database), your app machines might have the following process topography:
- App server 1: 3 web (thin start), 3 workers (rake jobs:work), 1 clock (clockwork clock.rb)
- App server 2: 3 web (thin start), 3 workers (rake jobs:work)
You should use Monit, God, Upstart, or Inittab to keep your clock process running the same way you keep your web and workers running.
@mamccr posted the following example to use this with daemons gem and rc.d script.
clockwork_control.rb
require 'daemons'
require 'clockwork'
clock_path = File.join(File.expand_path(File.dirname(__FILE__)), 'clock.rb')
Daemons.run_proc('clockwork') do
STDERR.sync = STDOUT.sync = true
require clock_path
trap('INT') do
puts "\rExiting"
exit
end
Clockwork::run
end
clock.rb
require 'clockwork'
include Clockwork
# anything as you lie
Put them in the same directory, and start a daemon with
ruby clockwork_control.rb start
When using with rc.d, prepare a script like this.
cd YOUR_PROJECT_DIRECTORY && export PATH=$PATH:/usr/local/bin && ruby clockwork_control.rb ${1}
@mamccr's original example is for Rails(tomykaira/clockwork/#14).
Created by Adam Wiggins
Inspired by rufus-scheduler and resque-scehduler
Design assistance from Peter van Hardenberg and Matthew Soldo
Patches contributed by Mark McGranaghan and Lukáš Konarovský
Released under the MIT License: http://www.opensource.org/licenses/mit-license.php