Amazon spot instance deployment antipattern

I'm pretty sure I'm doing things the wrong way and I'd like to know the right way.

The problem I face is:

• I have a distributed batch job
• I use AWS spot instances to fulfill it
• always a one-off request (so I'm after simple solutions)
• The instances need some configuration and setup, and they need to be sent tasks to do

I typically solve it like this:

• Bring up one worker and configure it
• Save an image of it
• Then bring up all the other instances
• Then I'll do a big, annoying, copy paste job, to basically create a shell script like this: $HOST=123.123.123.1 $TASK=task1 ssh ubuntu@$HOST task_runner $TASK & $HOST=123.123.123.2 $TASK=task2 ssh ubuntu@$HOST task_runner $TASK & $HOST=123.123.123.3 $TASK=task3 ssh ubuntu@$HOST task_runner $TASK & ...

I'd like to do this better, so I'm after ways to make it easier to:

1. Set up AWS instances. Sometimes there's a boot-time code that needs to be run, and this means more big tasks like the one above
2. Get the instance list. No more copy-pasting!
3. Run all the tasks on all the instances.
4. Shut them down when I'm done

Ideally I'd like to cover the whole lifecycle, end to end.

Lastly, I want something pretty simple. Because it's a one off task, I don't really want to have to deal with a whole lot of config options and wiring.