Would be great to see an example of how to override the worker's name.

Under normal conditions delayed jobs run and report results as expected:

*** Starting job worker host:Erics-MacBook.local pid:19112
About to perform job
1 jobs processed at 6.1079 j/s, 0 failed ...

However, when a job fails, it attempts to perform it a second time and displays the result as 2 errors:

*** Starting job worker host:Erics-MacBook.local pid:19103
About to perform job
About to perform job
2 jobs processed at 22.1354 j/s, 2 failed ...
About to perform job
About to perform job
2 jobs processed at 89.2026 j/s, 2 failed ...
2 jobs processed at 47.0045 j/s, 2 failed ...
About to perform job
About to perform job
2 jobs processed at 90.6083 j/s, 2 failed ...

Here is the code that performs the delayed job:

class TestJob
  def initialize
    puts "Job created"
  end

  def perform
    puts "About to perform job"
    raise "Ooops"
  end
end

When MAX_ATTEMPTS is set to 0 and an error occurs the last_error field is not set.

[sorry, this was my fault]

Hi,

the model, my jobs get their info from is stored in a couchdb instance.

The background process produces these kind of errors:

MailJob failed with ActiveRecord::RecordNotFound: Couldn't find MailReceiver with ID=b8a6908bd4fdbf4b86236f1e44ff1056 - 1 failed attempts

But sometimes the queuedelivery works with the second or third try!
I could imagine that DJ expects the MailJob model to come from AR?

Any hints are very welcome ;)
-- Frank

If you're using handle_asynchronously, I think it's reasonable to assume that you want the delaying behaviour to be "transparent" within your app. Therefore, in test mode, the relevant methods should not actually become a delayed job.

I'm achieving this currently using:

# Disable transparent delayed_job methods in test mode
module Delayed::MessageSending
  def send_later(method, *args)
    send(method, *args)
  end
end

This has the side effect of also disabling send_later. I have mixed feelings about whether that should actually happen, as it's a more explicit call, so would be interested in thoughts on that.

In textile, you need to use pre and code tags around large code chunks.

I've made changes to my to display more nicely: http://github.com/technicalpickles/delayed_job/tree/readme_code

It seems that a user of delayed_job is always going to need to make a migration for the delayed_jobs table. It would make sense to provide a generator to automate this process.

I've made a small one in my fork: http://github.com/technicalpickles/delayed_job/tree/generator

As you can see, the daemon uses the wrong db credentials unless RAILS_ENV is passed in:

[root@my_server ~]# /home/deploy/my_app/script/delayed_job -e staging start
/home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/mysql_adapter.rb:547:in `real_connect': Can't connect to local MySQL server through socket '/tmp/mysql.sock' (2) (Mysql::Error)
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/mysql_adapter.rb:547:in `connect'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/mysql_adapter.rb:199:in `initialize'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/mysql_adapter.rb:74:in `new'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/mysql_adapter.rb:74:in `mysql_connection'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/abstract/connection_pool.rb:223:in `send'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/abstract/connection_pool.rb:223:in `new_connection'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/abstract/connection_pool.rb:245:in `checkout_new_connection'
    from /home/deploy/rails_apps/my_app/releases/20090916055509/vendor/rails/activerecord/lib/active_record/connection_adapters/abstract/connection_pool.rb:188:in `checkout'
     ... 19 levels...
    from /home/deploy/my_app/config/../vendor/rails/railties/lib/initializer.rb:112:in `run'
    from /home/deploy/my_app/config/environment.rb:6
    from /home/deploy/my_app/script/delayed_job:3:in `require'
    from /home/deploy/my_app/script/delayed_job:3
[root@my_server ~]# RAILS_ENV=staging /home/deploy/my_app/script/delayed_job -e staging start
[root@my_server ~]# 

Received this in our delayed_job.log... and then the DJ process dies. My best guess so far is there was a momentary network disruption between firewalls of our app and db networks.

<ActiveRecord::StatementInvalid: Mysql::Error: MySQL server has gone away: UPDATE delayed_jobs SET locked_by = null, locked_at = null WHERE (locked_by = 'delayed_job host:myhost pid:12345') >

Would be nice if DJ had more robust connection handling... i.e. catch this issue, and retry the connection until it is successful... otherwise we have to write a script to restart DJ periodically, or detect when jobs in the queue are stale.

Delayed::Job

def self.db_time_now
  (ActiveRecord::Base.default_timezone == :utc) ? Time.now.utc : Time.zone.now
end

App is in rails 2.2.3 on ruby 1.8.7.

[apphome]/config/environments does not set config.time_zone.

error demo'd in console:

> ruby script/console 
Loading development environment (Rails 2.2.3)
>> ActiveRecord::Base.default_timezone
=> :local
>> Time.zone.now
NoMethodError: You have a nil object when you didn't expect it!
The error occurred while evaluating nil.now
        from (irb):2

Simple work-around: use Time.now instead of Time.zone.now, which works for my Rails settings, but not sure that is a general solution.

We were working with some beefy jobs that silently overflowed the default TEXT columns in mysql. The truncated YAML data were unable to be loaded, and were causing the daemon to crash. It looks like incomplete error handling in the job.deserialize method to me; the second YAML.load call doesn't have a rescue block, and the main rescue block for the method doesn't catch the YAML syntax error.

Periodically, I will have jobs that should be running, but... aren't. I mean they have values in the "locked_at" and "locked_by" columns, and they are clogging up the queue, but they never finish executing.

Run "script/delayed_job start" twice, two processes will get started. The oldest one has to be killed manually since the pid file gets overwritten. The start action should check and not start a new process if one is already running (pid file exists, and a process with that pid is also running).

Perhaps the Delayed::Job queue is order agnostic, but I would expect it to process jobs of the same priority in the same order as they were added to the queue.

Here's a failing test that I wrote against one of my models:

def test_delayed_job

comment = Comment.create!(:body => 'This is a comment')
assert_equal 0, Delayed::Job.find(:all).size
comment.send_later(:update_attribute, :body, 'Comment the first')
comment.send_later(:update_attribute, :body, 'Comment the second')
assert_equal [2, 0], Delayed::Job.work_off
assert_equal 'Comment the second', comment.reload.body

end

Unfortunately, that test fails with the following output:

<"Comment the second"> expected but was <"Comment the first">.

The order in which the two updates are dequeued differs from the order in which the two updates were enqueued. Is this the expected behavior?

Since github doesn't support compiling gems anymore, do you have any plans to move delayed_job onto gemcutter? I'm trying to get rid of my references to github as a source, but delayed_job is one of the few stragglers left.

I can push jerryvos-delayed_job to gemcutter for my own purposes, but would prefer to just use a real delayed_job gem or a tobi- one.

I'd happily provide any assistance in doing this and/or push a gem there myself, but don't want to steal ownership of delayed_job out there (though if you ask me to push it I can and then add you as an admin of it).