zero-downtime restarts

[nevans]

for those time-sensitive mission critical jobs that just can't be delayed by 15 seconds without annoying users (or those apps that take more than 15 seconds to boot up).

[lwoodson]

We have done this at my company by not using a pid file and having a before first prefork hook that will send INT to any resque pool instances not managing the process that is invoking the code. So the old master process dies immediately when the new workers are ready to work. Old workers get to finish and then die off. If you want, we could port to a pull request as a baked in feature of resque-pool. Thoughts?

[jjulian]

I like that feature. +1 for a PR.

[lwoodson]

Okay, we'll put something together.

[nevans]

FYI: I'm hoping #132 will be merged for the next release, so I'm closing this ticket in favor of that one.