This is an overly simplistic example of how we could use very basic reflection combined with a trait for any process to be able to handle deployment. The idea here is that we have this common entry point with a main that takes as arguments a fully-qualified class name and then the array of other arguments. The first argument must be the object you want to run everything else will be specific to that object.
The object must use the DataProcess trait which defines an execute method.
The execute method will receive an Array[String] and do what it wants with them.
This combined with an executable jar (either through java -jar runner.jar) or by pre-pending a shell script onto the jar. We can debate the merits of this.
My suggestion would be that this jar could also contain a common logback.xml and any other common resources (I'll get in trouble here but graphite config, sentry config, etc).
The end result is the ability to do something like:
./runner dataplant.FIXLogLoader --date 20150400
I have to work out the details of using this to run classes that aren't apart of runner.jar but on the surface I appreciate this approach for its simplicity of deployment as well as the uniformity that data processes will have by adhering to the DataProcess trait. Finally, it begins the unifying of common resources (like logging, etc)
Its not perfect but its a good idea that needs some vetting.