Resque scheduler on heroku without extra workers

require 'resque/tasks'
require 'resque_scheduler/tasks'
# Scheduler needs very little cpu, just start it with a worker.
desc "schedule and work, so we only need 1 dyno"
task :schedule_and_work do
  if Process.fork
    sh "rake environment resque:work"
  else
    sh "rake resque:scheduler"
    Process.wait
  end
end

15 thoughts on “Resque scheduler on heroku without extra workers”

Adam Perkins says:

2012-09-20 at 17:36:03

This is interesting and thanks for sharing. I’m newish to scheduling jobs on Heroku. Where does this set of code sit? Which file?

Thanks!

Reply
pragmatig says:

2012-09-20 at 17:38:11

e.g. Rakefile or lib/tasks/heroku.rake

Reply
David Fisher says:

2012-12-31 at 16:10:34

I have a the following code as well. Do I need it still with this?

task :work do
ENV[‘QUEUE’] = ‘*’
end

Reply
- David Fisher says:
  
  2012-12-31 at 16:11:41
  
  Nevermind, I obviously do. Since this just launches the work task OR the scheduler.
  
  Reply
David Fisher says:

2012-12-31 at 16:15:25

If I launch more workers- does this launch that many more schedulers? Or will it always keep only one scheduler up?

Reply
pragmatig says:

2012-12-31 at 16:19:48

It starts multiple, not sure if that is safe (does it schedule a job multiple times ?), also maybe schedulers die if there already is a scheduler running,
so you could make this more complicated by cheching if a scheduler is running or not, or check/make sure the multiple schedulers are safe to run

Reply
Chris Hough says:

2017-01-08 at 20:06:23

Would this impact this setup? http://stackoverflow.com/questions/41537142/heroku-procfile-resque-puma-rails-5-resque-scheduler-setup

Reply
- pragmatig says:
  
  2017-01-08 at 20:31:02
  
  Yes, you would only have the worker line (as rake schedule_and_work) and not the scheduler line.
  
  Reply
  - Chris Hough says:
    
    2017-01-08 at 22:16:43
    
    So this “worker: bundle exec rake resque:work COUNT=1 QUEUE=*” would change to “worker: bundle exe rake schedule_and_work COUNT=1 QUEUE=*” ?
    
    Would you be willing to add this to stack overflow too, I want to make sure you get credit for this? Also, I assume I delete the scheduler reference in the procfile correct?
  - Chris Hough says:
    
    2017-01-08 at 22:31:04
    
    Thank you for your help btw, I am trying to find 2 options. One now, and one for growth so I want to make sure I have the setup in place.
pragmatig says:

2017-01-09 at 1:32:42

Yep, that would work … remove scheduler and later when you scale you can split these 2 up again.

Reply
- Chris Hough says:
  
  2017-01-09 at 1:50:18
  
  I just updated the SO post. Can you confirm that matches the format you use. I greatly appreciate this as I was battling this for awhile now.
  
  Reply
- Chris Hough says:
  
  2017-01-09 at 4:48:32
  
  Thank you for checking the SO post. Did you want to leave an answer for credit?
  
  Reply
pragmatig says:

2017-01-09 at 5:00:23

Nah, got enough imaginary internet points 😉

Reply
- Chris Hough (@chrishough) says:
  
  2017-01-09 at 6:01:43
  
  Sounds good. Thank you so much for helping me with that. I appreciate it.
  
  Reply