workers don't fork for each job anymore

[databus23]

In qless 0.9.x it used to be the case that each job is processed in a separate fork from the parent.
The README still states:

The worker forks a child process for each job in order to provide resilience against memory leaks

Looks to me like this changed in current master. The SerialWorker which is also used by the ForkingWorker is processing jobs in a loop.
Was this change made deliberately? For what reasons?
What are the suggestions for getting back the old behavior and always have a fresh/clean process when working a job?

[brianchhun]

If I recall correctly, for forking workers, each worker is still forked from the parent process. However, instead of exiting after it completes, it will persist and continue picking up new jobs. This is problematic if you have jobs that don't clean up properly e.g., ones that continue hogging memory even after they complete.

To restore the old behavior, we added the following middleware into the worker:

module Qless
  module StopWorkerMiddleware
    def around_perform(job)
      super
      Rails.logger.info("Job #{job.jid} complete, shutting down child worker process #{Process.pid}")
      stop!
    end
  end
end

Qless::Workers::BaseWorker.class_eval do
  include Qless::StopWorkerMiddleware
end