How to cancel scheduled job with delayed_job in Rails?

Question

I am scheduling a job to run in say, 10 minutes. How to properly cancel this particular job without using any kind of dirty extra fields in model and so on. Is there any call to remove particular job, or jobs related to specific model, instance, etc?

Lonni

Answer 1

disclaimer: I am not an expert user of delayed_job...

"Is there any call to remove particular job, or jobs related to specific model, instance, etc?"

Delayed::Job is just an ActiveRecord object so you can find and destroy any of those records. Depending on your use case this could be handled different ways. If someone is going to manually destroy them this could be handled through an admin interface in your web app.

# list all jobs
Delayed::Job.all
# find a job by id
job = Delayed::Job.find(params[:id])
# delete it
job.delete

if you need some out of process task deleting jobs by 'job type' you could loop through each one and delete it if it matches your job; try this in script/console

class MyJob < Struct.new(:some_value);
    def perform
        # ...
    end
end

my_job = MyJob.new('xyz')
job = Delayed::Job.enqueue(my_job, 0, 1.hour.from_now)
job.name
# => "MyJob"
job.handler
# => "--- !ruby/struct:MyJob \nsome_value: xyz\n"

so given the above if you wanted to delete all jobs of type MyJob

Delayed::Job.all.each do |job|
    if job.name == "MyJob" then
        job.delete
    end
end

this may or may not help for your situation? in many cases you might want to delete a MyJob but only where the :some_value attribute was 'abc' and not 'xyz'. In this case you might need to implement a 'display_name' on your MyJob object. job.name will use this if it exists

class MyJob < Struct.new(:user_id);
    def perform
        # ...
    end

    def display_name
        return "MyJob-User-#{user_id}"
    end 
end

# store reference to a User
my_job = MyJob.new(User.first.id) # users.id is 1
job = Delayed::Job.enqueue(my_job, 0, 1.hour.from_now)
job.name
# => "MyJob-User-1"
job.handler
# => "--- !ruby/struct:MyJob \nuser_id: 1\n"

this way you could be more selective about which records to delete?

hopefully this gives you enough information on possible ways to handle it?

Arthralgia

You should probably use "destroy" instead of "delete". – Groundage

In order to loop over all the delayed jobs you need to do Delayed::Job.all.each do |job|. (Need the .each). – Augmentation

as far as I know this won't stop/cancel a job that is already being performed, right? – Nutgall

@PereJoanMartorell - you are correct, this will not cancel a job already running, the only way I know how to do that is to stop the DJ process and then delete the job record. – Arthralgia

Answer 2

delayed_job 3 introduced a queue attribute. This can be hijacked to schedule a cancelable job.

class MyJob < Struct.new(:user_id)
  def self.queue_name
    "something-unique"
  end

  def perform
    # ...
  end
end

#scheduler
my_job = MyJob.new(User.first.id)
#'cancel' pending jobs first
Delayed::Job.where(queue: my_job.class.queue_name).destroy_all
#queue it up
Delayed::Job.enqueue(my_job,
  queue: my_job.class.queue_name,
  run_at: 1.hour.from_now
)

Unexpressed

I would really not recommend doing things this way. Thequeue is meant to allow the app to scale. You've already gone through with a custom Struct, just pick a different attribute name house9 did. – Circumcise

Answer 3

You may also consider using delayed job's payload_object method, if you're looking for a job through its passed parameter only.

Delayed::Job.all.each do |job|
  job.destroy if job_corresponds_to_target?(job, target)
end

def job_corresponds_to_target?(job, target)
  job.payload_object.args.first == target.id
end

This simplist example do not use fully the payload_object returned:

=> #<Delayed::PerformableMethod:0x0056551eae3660 @object=ReminderMailJob, @method_name=:perform_later, @args=[3]> 

Raincoat

Answer 4

I think it may get pricey to loop through all queued jobs serialized field (:handler), especially when queue is large (for example when rails eventstore replays events that you've subscribed to to schedule a job).

So the solution that seems to work for me, avoiding surgeries, looks like this:

# some_specific_event_handler.rb or policy 

record_uuid = SomeModel.find(event.data[:id]).uuid
queue_name = "#{record_uuid}_update_notification"
Delayed::Job.where(queue: queue_name).destroy_all
UpdateNotificationJob.set(
  wait: 30.minutes,
  queue: queue_name,
).perform_later(record_uuid)

Delayed::Job is a Delayed::Backend::ActiveRecord

The :queue field is just any string. I don't think its value plays any role in when and how job will be executed unless your code does something with it.

So I hooked my app logic to that :queue field value and it worked for my case, where, by requirements:

• if new event was emitted then schedule a job
• if there had been any same class jobs scheduled for this event, then dump those.

Dupuis