How does max_attempts interface with ActiveJob retry_on method?

[jeremygottfried]

https://apidock.com/rails/v6.1.3.1/ActiveJob/Exceptions/ClassMethods/retry_on

The ActiveJob retry_on hook allows you to customize how you respond to specific errors, but it's unclear how that integrates with max_attempts in this gem. Does this gem keep a separate queue and ignore ActiveJob behavior, or is max_attempts just setting a global retry_on to catch all errors?

[shanecav84]

I think that AJ's retry_on supersedes DJ's retry logic such that AJ will handle further retries (the attempts argument of retry_on) and deletion. If AJ somehow fails to do any of that and raises an exception, the exception will bubble up to DJ and DJ's retry logic will kick in.

[shanecav84]

I've been using AJ but relying on DJ's retry logic by using rescue_from instead of retry_on to handle job exceptions then rethrowing to have that error bubble up to DJ to handle retries.

[cnadeau]

In case it might help someone:

I had the same problem trying to configure max_attempts on an ActiveJob and ended up basing some work out of https://github.com/rails/rails/issues/39961 to get it working

# Idea taken from : https://github.com/rails/rails/issues/39961
# Original code from https://github.com/rails/rails/blob/main/activejob/lib/active_job/queue_adapters/delayed_job_adapter.rb branch v6.1.7

class CustomDelayedJobAdapter < ActiveJob::QueueAdapters::DelayedJobAdapter
  def enqueue(job) #:nodoc:
    delayed_job = Delayed::Job.enqueue(CustomJobWrapper.new(job), queue: job.queue_name, priority: job.priority)
    job.provider_job_id = delayed_job.id
    delayed_job
  end

  def enqueue_at(job, timestamp) #:nodoc:
    delayed_job = Delayed::Job.enqueue(CustomJobWrapper.new(job), queue: job.queue_name, priority: job.priority, run_at: Time.at(timestamp))
    job.provider_job_id = delayed_job.id
    delayed_job
  end

  class CustomJobWrapper < ActiveJob::QueueAdapters::DelayedJobAdapter::JobWrapper

    # use the job instance's max_attempts property and serialize it using ActiveJob's which saves it within DelayedJobs.handler field as YAML
    attr_accessor(:max_attempts)

    def initialize(job)
      super(job.serialize)
      self.max_attempts = job.respond_to?(:max_attempts) ? job.max_attempts : nil
    end
  end
end

then use it in the Application's configuration

 config.active_job.queue_adapter = ::CustomDelayedJobAdapter.new()

and add the a max_attempts attribute on the desired ActiveJob

module Matchmaking
  class SingleRunJob < ActiveJob::Base
    attr_accessor(:max_attempts)

    def initialize(*args)
      super(*args)
      self.max_attempts = 1
    end

    def perform()
      # ...
    end
  end
end