canvas-lms/config/initializers/delayed_job.rb

config = {
  :backend => 'active_record',
}.merge((Setting.from_config('delayed_jobs') || {}).symbolize_keys)

case config[:backend]
when 'active_record'
  Delayed::Job = Delayed::Backend::ActiveRecord::Job
when 'redis'
  if Rails.env.production?
    raise "Redis Jobs are not yet ready for production"
  end
  Delayed::Job = Delayed::Backend::Redis::Job
  Delayed::Backend::Redis::Job.redis = if config[:redis]
    Canvas.redis_from_config(config[:redis])
  else
    Canvas.redis
  end
else
  raise "Unknown Delayed Jobs backend: `#{config[:backend]}`"
end

# If there is a sub-hash under the 'queue' key for the database config, use that
# as the connection for the job queue. The migration that creates the
# delayed_jobs table is smart enough to use this connection as well.
db_queue_config = ActiveRecord::Base.configurations[Rails.env]['queue']
if db_queue_config
  Delayed::Backend::ActiveRecord::Job.establish_connection(db_queue_config)
end

# We don't want to keep around max_attempts failed jobs that failed because the
# underlying AR object was destroyed.
Delayed::Worker.on_max_failures = proc do |job, err|
  if err.is_a?(Delayed::Backend::RecordNotFound)
    return true
  end

  # by default, keep failed jobs around for investigation
  false
end