rescue exceptions and run job.fail in perform

[mcfiredrill]

I needed to test some of my resque failure code, but I realized that Resque::Job#fail is not actually run if you are using inline. I think using ResqueSpec.inline simplifies testing quite a bit, so I wanted to use it, but I still needed failures to be reported. So I wrote this patch. I'm not sure if this fits everyones use case, or if this is even the right way to go about it.

Let me know what you think.

[mcfiredrill]

Oops, well there is definitely at least one problem. It seems that on_failure_hooks are running twice now.

[leshill]

Hi @mcfiredrill,

An alternative approach would be to test fail directly, something like:

  describe "MyJob.fail" do
    it "updates the fail count" do\
      expect do
        MyJob.fail
      end.to change(MyJob, :fail_count)
    end
  end

[mcfiredrill]

I changed my test slightly and fixed duplicate failure hook bug by upgrading resque. Let me know what you think, although the tests still aren't passing on travis for some reason.

[mcfiredrill]

Ah ok, so I'm actually relying on redis running for that test, I didn't realize that. I'm not sure yet how to test it without relying on redis.

[leshill]

Hi @mcfiredrill,

Have you overridden the fail code in resque? I suspect the default behavior is using redis. Perhaps you should mock out fail and set the expectation that way?

[mcfiredrill]

Nah I'm not overriding it, I'm just calling job.fail directly, like worker.rb does here:
https://github.com/resque/resque/blob/1-x-stable/lib/resque/worker.rb#L235
https://github.com/resque/resque/blob/1-x-stable/lib/resque/job.rb#L291

Mocking out fail sounds like a good idea, I'll try that out.

[mcfiredrill]

So I did that, let me know what you think. Maybe I'm wary of using any_instance but I'm not sure how to get the specific instance, maybe if enqueue returned the job instance...

[leshill]

Hi @mcfiredrill,

Hmmm. Looking at that, it is not exactly what we want. Job#fail is what has the Redis dependency, and stubbing it in the specs still leaves the dependency at runtime.

We want a test double at runtime for Job#fail. If you want to try to write it, please do. If not I will be doing something like this in perform:

def perform(queue_name, payload)
  job = new_job(queue_name, payload)
  begin
    job.perform
  rescue Object => e
    fail_job(job, e)
  end
end

def fail_job(job, e)
  # run failure hooks here
end

Make sense?

[mcfiredrill]

Alright, I guess I kind of missed the point there. I'm actually testing a custom failure backend via this, which I guess I shouldn't be doing, I should be doing an integration test with full resque. I can still set inline manually without resque_spec.

I'm still having fun hacking on this project though, so I think I'll attempt to write this.

[kyrylo]

Any news/plans on this? At the moment I'm using a fork of this gem with this PR merged in.

[chrishough]

bump, I am tracking this too. I am trying to test job retries right now and can not seem to get around this :(

[kyrylo]

@chrishough try this https://github.com/airbrake/resque_spec

[chrishough]

@kyrylo are you referring to this ::

When /the (\w+) queue runs/ do |queue_name|
  ResqueSpec.perform_all(queue_name)
end

in that gem? do also use Resque.inline = true or keep that false?

[kyrylo]

It's just a fork where this PR is merged. You don't want to touch Resque.inline, if I recall correctly. Unfortunately, I don't test job retries, but it works quite well for testing normal jobs.

[leshill]

Hi all,

Thanks for following up, this project needs a new maintainer (#122). Please drop a note if interested.

Regards.