Running workers on a seperarte machine about bccresquebundle

[epicwhale]

Deploy your app source code to the multiple servers and run the workers there? As long as the REDIS instance is centralized, I don't see the issue?

[malisetti]

Thanks. I haven't tried anything yet to emulate this. I wanted to know the way.

[epicwhale]

Should work without hassles. Don't see a reason not to.

Let's close this issue.

[ruudk]

Or, If you don't like to deploy the web app to the separate server. Just setup PHP Resque yourself on the other server. Connect it to the same Redis instance and your done.

[danhunsaker]

The trouble there is that you still need the job classes on the (second) worker instance. Doable, but you need to be mindful of the requirement.

[epicwhale]

Doesn't make sense to do partial deployments... The Job classes are all in your symfony2 app. Nothing is going to work fine without it.

[ruudk]

I was just pointing out that it's not a requirement to get it working :)

[danhunsaker]

Strictly speaking, yes, Resque itself will operate just fine without any other code. It won't, however, be able to properly execute any of the jobs without also including the job classes. If your project is designed around this factor, meaning your job classes are implemented separately from the non-job portions of your project, you can safely deploy these separate parts of your project without needing to include the full project in both locations. Most applications aren't designed with this in mind, however, for numerous reasons - not the least of which is code deduplication. It all depends on your project design, which in turn depends on development priorities (among other requirements).

For most users, you'll simply deploy the full project to separate servers, and configure one as your interface/job queueing server, with the other running only your workers. (The hosting service Pagoda Box uses this approach, as one example of how this might work in practice.)

[malisetti]

Thanks guys, you are the best. I will try to implement some of these practices.

[malisetti]

I was trying to use resque from php(consumer) and node js(producer used https://github.com/technoweenie/coffee-resque). I see the way these two are using redis is not similar to push jobs and to poll jobs. So there is no interoperability. I am wrong. I just moved to Gearman after this.

[danhunsaker]

Ah. That. Yeah. Coffee Resque is probably keeping up with the original Ruby version of Resque, where the PHP version is not. We're hoping to fix that, because interoperability between languages is actually one of the project goals, but progress is slow until the current (original) maintainer can find enough time to properly consider adding more maintainers.

Sorry we weren't ready in time for your project, but hopefully you'll keep an eye on us and give Resque another shot when we finally catch up! Best of luck!

[malisetti]

But the https://github.com/chrisboulton/php-resque says it's a direct port of ruby resque, then why the implementation is different ? @danhunsaker

[danhunsaker]

Because Ruby Resque is several versions further along. PHP Resque was ported a few years ago and hasn't been kept in sync, largely due to the original developer not having enough time.

Also, PHP and Ruby work differently, so there will be some differences because of that. Those aren't the kinds of differences that cause problems, though. The problems are coming from one implementation continuing to make changes and improvements, while the other implementation stays essentially the same as it was when it was ported originally.

[epicwhale]

@danhunsaker this issue can be closed as the worker.root_dir config is an intermediate solution to this problem.

[danhunsaker]

Not sure why I was tagged on that...