I wrote a script to automatically launch/terminate Mesos slaves on EC2 based on available resources: https://github.com/thefactory/autoscale-python/blob/master/examples/mesos_ec2.py
Scaling actions are performed by a provider-specific component, so supporting Google Compute Engine or another provider is simply a matter of implementing the same (trivial) interface. Same goes for incorporating other data sources or custom scaling logic: just write a drop-in replacement for an existing component. I've started a Python library around this pattern. Current support is...humble.. (only "Mesos based on disk/mem/cpu free on EC2"), but I'd love to build out a suite of components to handle different use cases (non-Mesos; non-EC2; alternative logic; etc). Please check it out: https://github.com/thefactory/autoscale-python Thanks to Connor Doyle for the early design feedback. If anybody else has thoughts on how to improve it, I'd love to hear them!
