Rufus::Scheduler.start required in production

[jmettraux]

Rufus::Scheduler.start required in production

[jmettraux]

Hello, I would need more details. Could you please gist your exact test script?

I don't have a Ubuntu 12.04, I'll try on a Debian system.

Thanks in advance.

[jamilbk]

It's probably something to do with how my unicorn+nginx setup is spawning processes, but the inner workings of Ruby threads are a bit over my head.

Here's the gist: git://gist.github.com/2584148.git

And thanks for the prompt reply. :-)

[jmettraux]

Well... I guess your Unicorn is spawning multiple Ruby runtimes. Let's say it's spawning 4 Ruby runtimes, at that point you'll have 4 schedulers.

Since the runtimes are spawns, the instances are copies, but the rufus-scheduler thread is not started, that's probably why you need to do "start". At some points your runtimes will get killed by Unicorn (new runtimes will get started), it depends on your configuration. When a runtime exits, its rufus-scheduler's jobs will get lost as well.

I'm guessing that your development environment is served by a plain Webrick while your production environment is Unicorn + Nginx.

I invite you to carefully read your Unicorn documentation, to learn about the Unix process model and to learn about Ruby threads. You cannot go on with such knowledge being over your head.

Please level up and/or look at alternatives.

I noticed you're using resque, why not take a look at https://github.com/bvandenbos/resque-scheduler ? It's a resque + rufus-scheduler integration.

Cheers,

John

[jamilbk]

Hey John,

Great feedback. Thanks.

I gave resque-scheduler a try but was having a hard time finding a good way to start the scheduler and keep it running with capistrano. Will keep up the fun