Not sure it is a sign of how wide spread the Amazon AWS or Hadoop are getting or that Clojure is becoming hotter and hotter, but this project sounds pretty cool:
Crane solves the problems for people running hadoop clusters; write all your config in clojure maps, and simply call (launch-hadoop-cluster config). You can run hadoop jobs from your local REPL, including the provisioning. All the configuration is in clojure. Automated shutdown for clusters when cascading jobs complete is on the way.
